Yoodli Prod

Three years ago, a company selling AI software to a large enterprise handed the account to a services team after the contract closed. Today, at Palantir, Anthropic, OpenAI, and a growing list of AI vendors, that engineer shows up months before the contract closes, sits in the room during the pitch, and stays embedded through the first stretch of deployment. That person carries the title forward-deployed engineer, and the shift in when and how they show up has changed what it takes to win enterprise AI deals.

The technical seller replaced the technical demo

For most of the SaaS era, sales engineering existed to answer questions in a demo and disappear once the deal closed. Implementation was someone else’s job, usually a professional services team that picked up the account weeks or months later. The forward-deployed engineer model breaks that handoff. FDEs get embedded early, often before a contract is signed, building working prototypes against a prospect’s actual data and actual workflows instead of a generic demo environment.

Palantir built its entire go-to-market motion around this idea long before “FDE” became a job title other companies borrowed. The pitch was never a slide deck. It was a working system, built on site, that solved a version of the customer’s real problem before the contract was signed. Anthropic and OpenAI have adopted a similar model for their own enterprise accounts, and the reason is direct: buyers of AI systems no longer trust a demo to predict what will happen against their own data and their own edge cases.

What winning an enterprise AI deal looks like now

Winning used to mean out-featuring a competitor on a spec sheet or beating them on price per seat. A buying committee evaluating an AI system now wants proof it will work against their data, their compliance constraints, and their existing tools before anyone signs off on a budget line and an executive sponsor. A vendor that can show a working prototype in the first meeting has already cleared a hurdle a slide deck cannot.

This changes the shape of the sales cycle. Buyers evaluate the FDE directly, alongside the product, assessing whether the vendor’s technical team understands their business well enough to build something specific to it, and whether that same team will still be there after the deal closes to make the deployment work. Research on the model from the Alexander Group frames this as a shift from selling a finished product to co-building a proven outcome with the buyer during the sales process itself.

Why this became a competitive advantage instead of a cost center

Pre-sales technical work has always cost money. What changed is where that cost sits on the P&L and what it buys. A company that treats forward deployment as overhead assigns generalist SEs to run demos and hands the hard problems to a services team after signature. A company that treats it as a differentiator hires senior engineers who can build and adjust real deployments in front of the customer, and puts them in the deal from the start.

The second approach costs more per deal. It also closes more of the deals that matter, because the buyers who evaluate enterprise AI carefully, security teams, procurement, technical stakeholders, are the ones who decide whether a six or seven figure contract gets signed. A vendor that can prove technical credibility inside the sales cycle skips months of trust-building that used to happen after signature. That speed compounds across a pipeline. It is hard for a competitor to copy quickly, because copying it means hiring and developing a different kind of person than a traditional AE or SE, and building a sales process that gives that person room to work before the deal closes.

The hiring math backs this up. Strong FDEs are scarce because the role asks for two skill sets that rarely sit in the same person: the engineering depth to build a working system under time pressure, and the presence to do it in front of a skeptical VP of Engineering or CISO who is deciding whether to trust the vendor with production data. Companies that figure out how to find, train, and retain that combination end up with a sales capability a competitor cannot buy off a job board in a quarter. That is the part of the FDE model that holds up as a durable advantage rather than a hiring trend. It takes years to build the internal muscle to identify these people, pair them correctly with the rest of the deal team, and give them enough reps that the pairing works under pressure instead of only in a rehearsal room.

The model breaks if the people around the FDE aren’t ready

An FDE who writes good code and designs a good architecture is not enough by itself. The FDE sits inside a sales motion with an account executive, a solutions consultant, and often a customer success lead, and the deal moves at the pace of the weakest handoff in that group. If the AE cannot speak accurately about what the FDE built, or the FDE cannot translate a technical constraint into an answer a VP will accept, the credibility the model is supposed to buy disappears in the room.

This is where a lot of companies scaling the FDE model run into trouble. They hire strong individual engineers and assume technical skill will carry the deal. It carries the first meeting. The eight or ten touchpoints between discovery and signature take more, since each one brings a different stakeholder asking a different version of the same skeptical question. Teams that handle this well rehearse those conversations before they happen, not just among the FDEs but across the whole deal team, so the AE, the FDE, and the CS lead give the buyer one consistent, credible story instead of three separate ones.

Yoodli built its AI roleplays for sales onboarding around exactly this gap: the difference between a team that knows the material and a team that has practiced saying it out loud, under the kind of pressure a real buyer applies. Standing up an FDE program takes more than a job posting for strong engineers. It takes an AE and an FDE who have rehearsed the handoff enough times that a skeptical question from a CISO does not derail the pitch.

What this means for building and training GTM teams

Companies that want the FDE model to work are rethinking who they hire and how they ramp them. The role sits between engineering and sales, and most engineers have never been trained to handle a room full of stakeholders who are evaluating them as much as the product. Most AEs have never had to co-present with someone writing code live. Onboarding for this hybrid role cannot be a shortened version of a normal SE ramp. It needs its own path.

The companies pulling ahead build that path on purpose. They pair new FDEs with AEs early. They rehearse the specific objections that come up in enterprise AI deals around data security, model reliability, and integration risk. They treat the deal team’s ability to present as one unit as a skill that gets coached, not assumed. Google Cloud has certified more than 15,000 people through structured sales training, RingCentral cut call-center certification time by 90 percent using AI roleplay practice instead of shadowing and manuals, and Harness cut the time its managers spend reviewing sales training by 75 percent.

Each shows what happens when a company stops treating rehearsal as optional and starts treating it as part of the operating model. The same logic applies to a forward-deployed engineering org. A technically strong FDE who has never rehearsed a live objection is a liability in a room where the buyer is testing composure as much as competence.

Model quality alone rarely decides an enterprise AI deal anymore. The vendors winning today have technical and commercial teams that show up together, prepared, and leave the buyer with proof instead of a promise. That is the real moat, and it gets built one rehearsed deal team at a time.

Most sales onboarding programs have an end date. A new rep finishes the product training, passes a certification, shadows a few calls, and graduates to a full quota. Then the formal learning stops. The product keeps changing, the pitch keeps changing, and the market keeps changing, but the training calendar goes quiet until the next kickoff.

Everboarding is the fix for that gap. This post covers what the term means, why enablement teams are adopting it, and how to run it without burying reps in more training.

What everboarding means

Everboarding is continuous onboarding. It treats learning as an ongoing part of the job instead of a phase that ends after a rep’s first 90 days. The same structure that gets a new hire ready (clear skills, practice, feedback, certification) keeps running for tenured reps as products, messaging, and buyers evolve.

The term is most common in sales and customer-facing teams, but the idea applies anywhere skills decay or change faster than annual training can keep up.

Why one-time onboarding falls short

Onboarding front-loads a large volume of information into a short window. New hires retain some of it, forget much of it, and learn the rest on live calls with real buyers. That last part is expensive, because every early mistake happens in front of a prospect.

Three things make the problem worse over time:

• Products and pricing change. A rep who onboarded a year ago learned a version of the product that no longer exists.
• Messaging changes. New positioning, new competitors, and new objections show up faster than a yearly training cycle.
• Skills fade without use. A rep who rarely handles a certain objection gets worse at it, even if they handled it well during onboarding.

Everboarding addresses all three by making short, regular practice and recertification part of the normal operating rhythm.

What an everboarding program includes

A working everboarding program usually has four parts.

1. A living skills map. Define the conversations each role needs to handle well: discovery, demo, objection handling, pricing, renewal. Revisit the list every quarter and add whatever the business has changed.

2. Short, frequent practice. Replace long annual sessions with small practice reps tied to real work. Ten minutes of practicing a new objection before a week of calls beats a two-hour workshop a rep will forget.

3. Recertification on change. When the pitch or product changes, recertify the team on the new version. Google Cloud has certified more than 15,000 people on its new GTM pitch with Yoodli.

4. Feedback loops with managers. Practice data should reach the manager, so 1:1s can focus on the specific gaps a rep is showing instead of general impressions.

How to run everboarding without training fatigue

The risk with continuous learning is that it feels like continuous homework. A few rules keep it useful.

Tie every assignment to something happening now. A new product launch, a competitor move, or a common objection from last month’s calls. Reps engage with practice that helps them on their next call.

Keep sessions short. Everboarding works best in small doses. If a practice assignment takes longer than a coffee break, split it.

Let reps practice privately. Many reps avoid live roleplays with peers or managers because mistakes feel public. AI roleplay gives them a place to try, fail, and retry before anyone is watching.

Measure skill over completion. A completion rate tells you who clicked through. A scored practice session tells you who can actually deliver the new message.

Where AI roleplay fits

Everboarding breaks down when it depends on managers and enablement teams running every practice session by hand. There are not enough hours for that at scale.

AI roleplay removes that constraint. Enablement teams build a scenario once, around their own methodology, content, and buyer personas, and every rep can practice it on demand. Each session produces consistent feedback and a score, which gives managers a clear view of who is ready and who needs coaching.

Speed matters here too. RingCentral reduced call-center certification time by 90% with Yoodli. When certification takes a fraction of the time, recertifying after every pitch change becomes realistic.

How to start

You do not need to rebuild onboarding to begin everboarding. Start small:

1. Pick one conversation that changed recently, such as a new product, a pricing update, or a new objection.
2. Build one practice scenario around it.
3. Ask every customer-facing rep to complete it and hit a passing score.
4. Review the results with managers and repeat next month with the next change.

After two or three cycles, the rhythm becomes part of how the team works. That rhythm is the program.

FAQ

Is everboarding the same as continuous learning? They overlap. Everboarding applies the structure of onboarding (defined skills, practice, and certification) to continuous learning, so it is more specific and easier to measure.

Who owns everboarding? Usually sales or revenue enablement, with frontline managers owning follow-through in 1:1s.

How often should reps recertify? Whenever a core conversation changes, plus a regular cadence (often quarterly) for the skills that matter most.

ADDIE is a five-phase framework that instructional designers and corporate learning and development teams use to build training programs. The name stands for Analysis, Design, Development, Implementation, and Evaluation, and each phase produces something the next phase depends on. Florida State University’s Center for Educational Technology built the original version of the model in the 1970s for military training, and it has stayed the default starting point for instructional design work since. If you work in L&D, sales enablement, or any function that builds training for other employees, you have probably used some version of ADDIE without calling it that.

Teams reach for a structured model like ADDIE because building training without one tends to produce the same failure pattern: someone builds a course based on what they assume reps need, ships it, and finds out months later that it never addressed the actual performance gap. ADDIE forces a team to define the problem and the success measure before anyone opens a slide deck or a script editor, then checks the work at each stage instead of only at the end.

To make the five phases concrete, picture a 400-person software company whose new sales reps take five months to reach full quota, compared with three months for reps hired two years earlier. The VP of sales enablement gets asked to fix ramp time. Here is how ADDIE would carry that project from a vague complaint to a shipped training program.

Analysis

The Analysis phase asks what problem the training is supposed to solve, and whether training is actually the right fix for it. This is where a team gathers evidence instead of guessing at content. In the ramp-time example, the enablement team interviews five sales managers, pulls win-loss data from the CRM, and listens to a sample of discovery calls from newer reps.

The interviews turn up a specific gap. Reps understand the product well by month two, but they freeze when a procurement lead pushes back on price or a legal reviewer raises a contract objection in real time. That distinction matters, because it points the rest of the project toward practice and away from another round of product training videos.

A good Analysis phase ends with a written problem statement and a target metric, the two things the rest of the project gets checked against. In this example, the statement is short: reps lose deals at the procurement stage because they have not practiced handling live pushback, and the target is closing that gap within one quarter. Everything built in the next three phases gets checked against that statement.

Design

Design turns the problem statement from Analysis into a blueprint. This phase sets learning objectives, decides how content will be sequenced, and defines how success will be measured before anyone builds a single asset. Skipping it is how teams end up with training that covers the wrong material in the wrong order.

For the ramp-time project, the enablement team decides reps need repeated practice against three objection types before their first live call with a procurement stakeholder: price pushback, contract stalling, and competitor comparisons. They storyboard what each practice scenario should cover and agree that success means a rep can handle all three objection types without a manager stepping in.

The output of Design is usually a planning document: objectives written in observable terms, a rough outline of each module or scenario, and an assessment plan describing how a manager or the system itself will judge whether a rep has actually learned the skill. A subject matter expert, often a top-performing rep or a sales director, usually reviews this outline before Development starts, which catches gaps while they still cost a paragraph to fix instead of a finished module.

Development

Development is where the blueprint becomes real material: scripts, slides, videos, job aids, or practice scenarios. This is usually the slowest phase in a traditional ADDIE project, because writing realistic dialogue, recording video, or building a full e-learning module can take weeks of work per hour of finished training. It is also the phase where most review cycles pile up, since a subject matter expert has to check each draft against how a real conversation actually goes.

This is also the phase where AI roleplay changes the math. Instead of scripting every branch of an objection-handling conversation by hand and waiting for a full course to be built before anyone can try it, a team can build a rough draft of a scenario from real call transcripts and sales scripts, test it the same week, and refine it against actual objections instead of guessing at dialogue up front. Yoodli has a walkthrough of how to train AI roleplay on your own sales content that covers what that build process looks like in practice.

Yoodli has also moved content creation into the roleplay builder itself. With auto-generated learning content, which launched in June 2026, an admin uploads existing decks, product briefs, or playbooks and gets structured learning content back in minutes. The admin can then edit it, regenerate sections, and export it as a PDF or PowerPoint. That work used to take weeks.

Implementation

Implementation is when the training goes live in front of the learners it was built for. A rollout plan usually covers who gets trained first, how instructors or managers are briefed, and what support exists if something in the material does not land.

For the objection-handling program, enablement pilots the three scenarios with one five-person pod before rolling them out to the full 40-person team. Because the scenarios are AI roleplays rather than a finished instructor-led module, the team can pilot in week two instead of week eight, and adjust a confusing scenario branch based on manager feedback without re-shooting video or rewriting a slide deck. Managers get a short briefing on how to review a rep’s scored practice sessions before the pilot pod starts, so coaching conversations reference the same data the reps are practicing against. Teams weighing a similar pilot can talk to Yoodli’s team about what a first scenario build and pilot typically involves.

Evaluation

Evaluation measures whether the training worked. For the ramp-time project, that means comparing the pilot pod’s time to full quota against a control group of reps who did not get the new practice scenarios, tracking manager scores on live objection-handling calls, and checking win rate at the procurement stage after 90 days.

In a mature ADDIE process, evaluation happens at the end of each phase, not only once at the very end of the project. A team checks whether Analysis correctly identified the gap, whether Design’s objectives still make sense once real learners see the material, and whether Development produced something Implementation can actually use. Waiting until the end of a multi-month project to find out the objectives were wrong is an expensive way to learn that.

Many L&D teams pair ADDIE’s Evaluation phase with Kirkpatrick’s four levels: how learners reacted to the training, what they actually learned, whether their behavior on the job changed, and what business result followed. For the ramp-time project, reaction shows up in pilot feedback, learning shows up in scenario scores, behavior shows up in manager observation of live calls, and the business result is the ramp-time number itself, checked again after a full quarter rather than guessed at from the pilot alone.

Waterfall in Theory, Iterative in Practice

Many teams run ADDIE as a strict waterfall. Analysis finishes and gets signed off, then Design starts and finishes and gets signed off, and so on through five stages that never reopen once closed. That version of the model creates long project timelines and locks a team into decisions made during Analysis long before anyone sees a working scenario or module. A stakeholder who only sees the finished product in week ten of a ten-week project has no real chance to catch a wrong assumption from week one.

The instructional design work behind ADDIE was iterative by intent, with each phase feeding information back into the ones before it. A team building the objection-handling program can develop one rough scenario as soon as Design produces a first objective, test it with two reps, and use what they learn to adjust their Analysis assumptions before building the rest. AI roleplay fits naturally into that loop, since a single scenario can be piloted before a full curriculum exists, which shortens the distance between Design, Development, and Implementation instead of treating them as three sealed boxes.

The Five Phases in Practice

ADDIE gives L&D and sales enablement teams a shared vocabulary for talking about training projects, whether a given team runs it as five strict stages or as a loop that revisits Analysis after every pilot. Analysis, Design, Development, Implementation, and Evaluation describe the order most training work naturally moves through, even when the lines between the phases blur once a project is actually underway. A team that names which phase it is in, and what that phase is supposed to produce, tends to ship training that matches the problem it set out to solve.

Teams comparing tools for the Development and Implementation phases can use Yoodli’s buyer’s checklist for AI roleplay platforms.

Most companies measure onboarding by counting who finished the course. It is the number every LMS surfaces on a dashboard without anyone asking for it, so it becomes the number leadership sees first. Completion tells you who showed up and clicked through the modules. It does not tell you whether a new rep can run a discovery call, handle a real objection, or close a deal without a manager coaching them through it live. If completion is the only metric in your onboarding report, you are measuring attendance and calling it competence.

Enablement and L&D teams that want a real answer to “is onboarding working” need metrics that track what happens after the course ends: how fast someone becomes productive, whether they actually hold the skills the course claimed to teach, what their manager thinks of their readiness, and how their early performance compares to reps who have been on the team longer. None of these show up in a completion report. All of them show up in the numbers that actually predict whether a new hire ramps or stalls.

Why Completion Rate Is the Wrong Headline Metric

Completion rate persists because it is easy. It comes straight out of the LMS with no extra instrumentation, no manager input, and no waiting period. A team can report 94% completion in the same week a cohort starts and look good doing it. The problem is that completion measures exposure to content, not retention of it or ability to apply it under pressure.

A rep can finish every module, pass a multiple-choice quiz on product features, and still freeze the first time a prospect pushes back on price. Watching a video about objection handling and handling an actual objection are different skills, and only one of them shows up in a completion percentage. Companies that stop at completion are optimizing for a number that correlates weakly, if at all, with the outcome they actually care about: reps who perform.

None of this means completion is worthless. Keep tracking it as a baseline prerequisite check, then build the rest of the scorecard on top of it.

Time-to-Productivity Is the Metric That Matters Most

Time-to-productivity, sometimes tracked as time-to-first-deal for sales roles, measures how long it takes a new hire to hit a defined performance bar: first closed deal, first successful support resolution without escalation, first fully independent client call. It is the closest single number to “onboarding worked” that most enablement teams have access to, because it ties training directly to a business outcome rather than a training-department outcome.

This number also exposes gaps that completion data hides entirely. Two cohorts can finish the same course at the same completion rate and land at wildly different time-to-productivity numbers if one cohort actually built skill during onboarding and the other memorized answers for a quiz. Structured practice built into onboarding shortens this window, because reps arrive at their first live call having already run the conversation multiple times against a scored practice partner instead of encountering the moment cold.

Certification Should Test Skills, Not Recall

Most certification quizzes test whether someone read the material. They ask which feature does what, or which step comes first in a five-step process. That is recall, and recall is the easiest thing to fake through cramming the night before a test. A rep can pass a knowledge quiz at 95% and still not know how to actually run the conversation the quiz was about.

A certification worth the name tests the skill itself. Can this person deliver the pitch, handle three predictable objections, and stay on message under a little pressure, evaluated against a rubric rather than a pass or fail toggle? RingCentral rebuilt certification around this kind of scored practice and cut call center certification time by 90%, because testing the actual skill instead of testing recall means fewer retakes and less time spent certifying people who were not ready the first time. When certification measures competence, the pass rate becomes a real signal instead of a formality everyone knows how to clear.

Manager Confidence Is a Metric, Not an Opinion

Enablement teams often treat manager sentiment as anecdotal color rather than data, which is a mistake. A structured manager confidence rating, collected on a consistent scale at 30, 60, and 90 days, is one of the fastest early-warning signals available. Managers see the new hire on live calls, in deal reviews, and in day-to-day interactions that no course completion record captures.

The key word is structured. “How do you feel about this rep” produces noise. “On a scale of 1 to 5, would you put this rep on a call with your most important prospect unsupervised” produces a comparable number you can track across cohorts and correlate against other metrics. Pairing that rating with objective coaching data from practice sessions gives managers something concrete to point to in the 1:1 instead of a gut feeling they cannot fully explain.

Compare New Hires to Tenured Reps, Not to a Fixed Target

A fixed onboarding target, like “hit 50% of quota by day 90,” gets outdated the moment your market, product, or comp plan shifts. A more durable comparison is new-hire performance trajectory against your own tenured reps at the same point in their ramp, using whatever performance metric matters for the role: quota attainment, resolution time, upsell rate.

If new hires are consistently closing the gap to tenured performance faster than last year’s cohort, onboarding is improving. If the gap holds steady or widens, something in the program is not transferring. This comparison also isolates onboarding’s actual effect, because it controls for market conditions that hit every rep, tenured or new, at the same time. A quota miss during a slow quarter should not read as an onboarding failure if tenured reps missed by the same margin.

Where Structured Practice Fits Into the Measurement

All four of these metrics point at the same underlying gap: the difference between exposure to knowledge and demonstrated ability to use it. Completion tracks exposure. Time-to-productivity, real certification, manager confidence, and trajectory against tenured reps all track ability, at different points in the timeline.

The reason ability is hard to measure early is that most onboarding programs have no place to observe it before a rep goes live with a real customer. Role-play with a manager takes manager time that most managers do not have to spare for every new hire. Shadowing shows a rep watching, not doing. AI roleplays close that gap by giving a new hire scored practice reps, on the actual scenarios they will face, before their first live call. Harness used this approach to cut sales-training review time by 75%, freeing manager time that used to go into listening to raw recordings and instead putting it into coaching on what the scores already flagged. That scored practice data becomes an early competence signal enablement teams can act on weeks before time-to-productivity numbers would otherwise surface a problem.

Building an Onboarding Scorecard That Reflects Reality

A scorecard built on these four metrics gives enablement a much more honest picture than a completion dashboard ever will. Completion stays on the sheet as a baseline check, because a rep who never touched the material is a real problem. But it sits alongside time-to-productivity, skill-based certification results, structured manager confidence scores, and trajectory against tenured performance, each pulling in signal that completion cannot provide on its own.

None of these metrics require exotic tooling. They require deciding, before the next cohort starts, what “productive” means for the role, building a certification rubric that tests the skill rather than the memory of a slide, putting a consistent confidence-rating cadence in front of managers, and pulling the same performance metric for new hires and tenured reps so the comparison is apples to apples. Teams that want to see how a scored-practice layer feeds these numbers directly can talk to the Yoodli team about what that looks like for their onboarding program.

A new sales rep’s first 90 days decide a lot. They set how quickly the rep reaches quota, how confident they feel on calls, and often whether they stay. Yet many onboarding plans are a list of training modules with dates attached, and very little that tells a rep or manager what “ready” looks like at each stage.

A 30-60-90 day plan fixes that by breaking ramp into three phases, each with clear goals and a way to check progress. This post walks through how to build one, with a template you can adapt.

What a 30-60-90 day sales onboarding plan is

A 30-60-90 day plan splits a new rep’s first three months into three stages:

• Days 1 to 30: Learn. Product, market, buyers, and process.
• Days 31 to 60: Practice. Apply that knowledge in simulated and supervised conversations.
• Days 61 to 90: Perform. Run real deals with increasing independence.

Each stage has specific outcomes, not just activities. The difference matters. “Attend product training” is an activity. “Can deliver the core product demo and pass a certification” is an outcome.

Days 1 to 30: Learn

The first month builds the foundation. The goal is for the rep to understand what they sell, who they sell it to, and how your team runs a deal.

Goals for this stage:

• Understand the ICP, key personas, and the problems your product solves for each
• Learn the sales methodology your team uses and the stages of your pipeline
• Know the top competitors and the most common objections
• Get comfortable with the CRM and core sales tools

How to check progress: Short knowledge checks at the end of each week, plus a first practice run of the elevator pitch and a discovery call opening. These early practice sessions are low stakes. They show the manager where the rep is starting from.

Common mistake: Loading the entire first month with content and no practice. Reps retain far more when they practice a concept soon after they learn it.

Days 31 to 60: Practice

Month two turns knowledge into skill. The rep should spend a large share of their time practicing the conversations they will soon have with real buyers.

Goals for this stage:

• Run a full discovery call against a realistic buyer persona
• Deliver the core demo and handle the top objections
• Shadow live calls and debrief them with a manager
• Start light prospecting under supervision

How to check progress: Certification on the core conversations. Set a clear passing bar for discovery, demo, and objection handling. A rep who passes is ready for live calls. A rep who does not knows exactly what to work on.

Where AI roleplay helps: Practice is the stage where most onboarding programs run out of capacity. Managers cannot roleplay every scenario with every new hire. AI roleplay lets reps practice realistic buyer conversations as many times as they need, with consistent feedback, and gives managers scores to review instead of sitting in on every session. RingCentral reduced call-center certification time by 90% with Yoodli.

Days 61 to 90: Perform

The final month shifts the rep onto live pipeline. Support stays in place, but the rep owns more of each deal.

Goals for this stage:

• Run discovery and demo calls independently
• Build and manage a starter pipeline
• Hit an early activity or pipeline target agreed with the manager
• Identify one or two skills to keep developing after onboarding

How to check progress: Review real call recordings with the manager, compare pipeline to the agreed targets, and look at practice scores on any skill that is still weak. Where a rep struggles on live calls, send them back to practice that specific conversation before the next one.

A simple 30-60-90 template

Stage	Focus	Key outcomes	How to measure
Days 1-30	Learn	Knows ICP, personas, methodology, competitors, tools	Weekly knowledge checks, first practice pitch
Days 31-60	Practice	Can run discovery, demo, and objection handling	Certification on core conversations
Days 61-90	Perform	Runs deals independently, builds pipeline	Call reviews, pipeline targets, practice scores

Adapt the outcomes to your motion. An SMB team with short cycles may compress this into 60 days. An enterprise team selling to buying committees may extend the practice stage.

Tips for making the plan stick

Write outcomes the rep can see. Share the plan on day one so the rep knows what “ready” means at each stage.

Give managers a weekly checkpoint. A short weekly review against the plan catches problems in week three instead of week ten.

Keep practice going after day 90. Onboarding should hand off to a regular practice rhythm, sometimes called everboarding, so skills keep up with product and messaging changes.

FAQ

How long should sales onboarding take? Most teams target 60 to 90 days of structured onboarding, though full ramp to quota often takes longer depending on deal size and cycle length.

Who owns the 30-60-90 plan? Enablement usually designs it. The hiring manager owns day-to-day execution and progress checks.

What should a rep be certified on before live calls? At minimum, the core pitch, discovery, and the most common objections.

Rolling out Claude across a company takes more than a license purchase and a welcome email. Employees try the tool once, get an average result, and go back to the way they worked before. What those teams skipped was practice: using Claude on real work with someone checking the output.

Yoodli now trains people on Claude the same way it trains sales teams on a pitch. An AI tutor walks each learner through the Claude interface, watches them work inside Claude over screen share, coaches their prompts in real time, and scores the result against a rubric your team defines. This post covers how that works, what it looks like in a live session, and how to set it up for your own rollout.

Why Claude Training Stalls After the Kickoff

The standard playbook is a wiki page, a prompt library, and a one-hour lunch-and-learn. Those teach people what Claude can do, and fluency only comes from using it. Reading about how to write a good prompt is passive. Opening Claude, writing a prompt for your own task, seeing a weak output, and fixing the instruction is active, and the skill only forms through the second kind of work.

Anthropic’s prompt engineering documentation is thorough, but reading it is closer to reading a cookbook than cooking a meal. Enablement and L&D teams already know this pattern from sales training. Reps don’t get good at objection handling by reading a battle card. They get good by running practice reps against realistic scenarios until the right response is automatic. Claude adoption works the same way, and Yoodli applies that same practice model to the tool itself.

Watch Yoodli Train Someone on Claude

In this short demo, Yoodli CEO Varun Puri goes through a Claude training session as a learner. The AI tutor opens by asking about his experience level. Varun says he is a novice who has used ChatGPT a little but doesn’t know much about Claude, and the tutor adjusts the session to that starting point.

From there the session runs in three parts:

• Interface walkthrough. The tutor shares its own screen and shows the Claude sidebar, the new chat button, and the input field. When Varun asks what Projects are, the tutor explains that they let you upload materials, set custom instructions, and group related conversations. It then covers Artifacts, the structured outputs Claude creates, such as documents, code, and websites.
• Hands-on task. The tutor stops sharing and gives Varun a real assignment: draft a LinkedIn post about how AI is improving his productivity. Varun shares his own screen with Claude open, and the tutor has him start a fresh chat so they can build the prompt together.
• Live prompt coaching. Varun’s first prompt is a one-liner asking Claude to create a LinkedIn post about AI productivity. The tutor calls it a solid start and pushes him to give Claude more context before he hits enter.

Behind the session, the tutor is scoring against a rubric called Prompt Writing Clarity and Effectiveness, which measures how clearly the learner writes prompts to get Claude to produce the intended work. The learner gets a personal tutor sitting next to them inside the real tool, and the admin gets a scored record of what that person can actually do.

How Claude Training Works in Yoodli

The demo runs on Yoodli’s AI platform certification capability. Four pieces make it work:

• A Tutor roleplay. Admins build the session in the Roleplay Builder using the Tutor template. The instructions define the greeting, the order of the interface walkthrough, the point where the learner takes over screen share, and the hands-on task.
• Visual content. Admins upload labeled screenshots of the Claude screens the tutor should show, such as the chat view, Projects, and Artifacts. The tutor presents them at the right moment in the walkthrough.
• Screen share inside the real tool. Learners practice in their own Claude workspace. The tutor sees what they type and what Claude returns, and coaches from there.
• Custom rubrics and Programs. Your team defines what good looks like for its own workflows, and the Claude sessions sit inside a Program so you can sequence them and track progress across the group.

The AI Tutor can also present your own training materials during the session, like an internal prompt library or a list of approved use cases, so learners can ask questions and apply the answer on the spot.

Build the Claude Training Around Real Work

The LinkedIn post in the demo is a good first exercise because everyone can relate to it. For an enterprise rollout, the hands-on tasks should come from each team’s actual week. Before you build anything, sit down with each function and list five to ten recurring tasks worth handing to Claude. Then turn the best ones into tutor sessions.

Team	Hands-on task in the tutor session	What the rubric checks
Sales enablement	Set up a Claude Project with the current battle cards and messaging doc, then draft a post-demo follow-up email	Custom instructions are specific, and the email uses approved positioning
Customer support	Draft a reply to an anonymized escalation ticket from the team’s own queue	The prompt includes tone, policy constraints, and the customer’s actual issue
RevOps	Turn a pipeline export into a summary Artifact for the weekly forecast call	The prompt defines the audience, the format, and the numbers that matter
L&D	Build a first draft of an onboarding module outline from existing training docs	The learner iterates on the output instead of accepting the first draft

Using the team’s own documents and language matters. A support lead who practices on a real ticket from their queue shows up on Monday able to do it again. A support lead who practiced on a vendor’s sample ticket usually doesn’t.

How to Set Up a Claude Tutor in Yoodli

Yoodli’s platform certification setup guide walks through the full process. The short version:

1. Define the learning outcome as a concrete task, such as “build a Claude Project for your territory and use it to draft an account plan.”
2. In the Roleplay Builder, create a new roleplay and select the Tutor template.
3. Write the instructions: greeting, interface walkthrough order, screen-share handoff, and the hands-on task.
4. Capture and label screenshots of the Claude screens the tutor will reference, then upload them as visual content with display instructions.
5. Configure the tutor’s persona, including its name, tone, and training style.
6. Add goals or a rubric so the session evaluates whether the learner completed the task well.
7. Add the roleplay to a Program so you can sequence it with other Claude sessions and track completion.
8. Run the full session yourself before assigning it to learners.

A practical sequence is three sessions: Claude basics and navigation, Projects with the team’s own materials, and a role-specific task scored against the rubric.

Certify on What People Can Do in Claude

Most AI tool training reports completion. Completion only tells you someone watched a video, and you still have no idea whether they can get a useful result out of Claude on their own. Yoodli certifies people on the task itself, scored against your rubric, inside the real tool.

This is the same certification model enterprise teams already run on Yoodli for other skills. Google Cloud used it to certify more than 15,000 people on a new go-to-market pitch, and Harness cut sales-training review time by 75% by moving assessment into Yoodli. The mechanics carry over to Claude directly: a clear standard, a realistic task, and scoring that doesn’t depend on a manager watching every rep.

The same Tutor setup works for ChatGPT, Microsoft Copilot, Gemini, or any other platform your company is rolling out, so enablement can run one certification motion across every AI tool in the stack.

Roll Out in Waves and Keep Managers in the Loop

Start with a pilot group of 15 to 30 people from one or two functions with repetitive use cases. Pick people who are already curious about Claude rather than assigning the pilot by seniority. Run them through the Claude Program over four to six weeks, then use their rubric scores and the prompts they struggled with to adjust the tutor sessions before the next wave.

Managers keep the habit alive after certification. Give them Yoodli’s rubric results for their team and one question to ask in every one-on-one: what did you use Claude for this week? A rep who knows their manager will ask has a reason to keep using the tool. Managers who use Claude visibly themselves do more for adoption than any mandate.

For measurement, skip seat utilization. Track certification pass rates by team, rubric scores by skill area, and whether people keep returning to Claude for the tasks they were trained on. Those numbers tell you whether the rollout changed how people work.

See Claude Training in Yoodli

If your company is rolling out Claude, the fastest path from licenses to fluency is practice inside the tool with coaching attached. Watch Varun’s Claude training session to see it in action, then talk to the Yoodli team about building a Claude certification program for your teams.

How to Train Financial Services Teams on Claude with Yoodli

Banks, insurers, and wealth management firms are rolling out Claude faster than their employees can learn it. The licenses get handed out, a kickoff deck goes around, and then adoption stalls because nobody practiced using the tool on real work. In a regulated industry that gap costs more than lost productivity. An advisor who never learned what belongs in a prompt will learn it on a live client file.

Yoodli closes that gap by teaching Claude inside Claude. An AI tutor walks each employee through the interface and assigns hands-on tasks built from your firm’s approved use cases. It watches them work over screen share and certifies them against your own rubric. This guide covers how that works and how to set it up for a financial services rollout.

Watch a Claude training session in Yoodli

Yoodli CEO Varun Puri recorded a short walkthrough of a live Claude training session. He plays a new user who has only tried ChatGPT a few times. The AI tutor asks about his experience level, then shares its screen to walk him through the sidebar, the new chat button, Projects, and Artifacts. It answers his questions as they come up. Then it gives him a real task, which is to draft a LinkedIn post about AI and productivity. Varun shares his own screen, opens a new chat in Claude, and types a first prompt. The tutor pushes him to give Claude more context before he runs it.

Any Claude workflow your team needs to learn can use the same session structure. For a financial services team, the task changes from a LinkedIn post to work your compliance team has approved.

Why a kickoff deck doesn’t teach anyone Claude

A slide deck can explain what Claude is and show a few example prompts. It can’t watch an analyst write a vague prompt and tell them how to fix it. It also can’t catch a relationship manager pasting an account number into a chat window. People learn Claude by using it on work that looks like their job, with someone correcting them as they go. Most rollouts skip that step because live coaching doesn’t scale across thousands of employees.

Anthropic’s getting started guide for Claude for Financial Services covers setup and features. Your enablement team still has to make sure each advisor, underwriter, and service rep uses the tool well and inside your firm’s rules.

How Yoodli teaches Claude

Four pieces make the training stick.

The tutor teaches before it tests. You upload your prompt libraries, approved use case list, and internal AI policy. The tutor uses them during the session, and learners can ask questions and apply the answer right away.

Practice happens in the real tool. Through screen share, Yoodli watches the learner work inside Claude and coaches them in the moment.

Your rubric decides what passing means. You set the criteria. One example is “Did the learner iterate on the output at least once?” A financial services team might add “Did the learner remove client names and account numbers before pasting the document?” or “Did the learner ask Claude to cite the section of the filing it used?”

Certification is based on doing the work. When a learner is ready, they complete real tasks in Claude while Yoodli evaluates them. Completing a module doesn’t count as certification.

The same approach works for ChatGPT, Copilot, and internal tools, so firms running more than one AI platform can train on all of them in one place.

Start with the use cases compliance approved

In most industries, enablement picks a few workflows and launches training. In financial services, compliance, legal, and information security review the specific tasks first. A wealth advisor drafting a client email carries a different risk from an underwriter summarizing a loan file, and both differ from a call center rep looking up policy language mid-call.

The review should produce three lists: approved use cases, use cases that need more scoping, and use cases that are off the table for now. That list becomes your Yoodli curriculum. Each approved use case becomes a tutor session, and each restriction becomes a rubric item. When compliance updates the list, you update the session, and employees certify again on the new version. Yoodli Programs track completion, so you can show which employees certified after each change.

Build a track for each role

One generic Claude module underserves everyone. Group sessions by function and build each track around the use cases that group is approved for.

Role	First Claude sessions to build	What the rubric checks
Wealth advisors	Drafting a client follow-up email for review, summarizing market commentary	No client identifiers in prompts, human review before sending, nothing that reads as individualized advice
Underwriters and credit analysts	Summarizing a loan file or public filing, comparing policy language	Data classification, asking Claude to cite sources, checking figures against the source
Contact center and service reps	Finding policy language, drafting internal case notes	What can be pasted during a call, escalating requests that fall outside approved use
Operations and finance	Reconciling internal reports, drafting internal memos	Approved source documents, verifying output before it is shared

Start narrow with drafting, research, and summarization. None of these ask an employee to make a compliance call on the spot, because the judgment happens before the task (which documents are approved) or after it (a person reviews before anything goes external).

Make data handling a practiced habit

“Don’t put sensitive data into the tool” doesn’t hold up under deadline pressure. Training has to name the categories that matter for your firm, such as account numbers, Social Security numbers, material non-public information, and client identifiers. It also has to give employees a clear path for gray areas.

Yoodli lets you practice this directly. Build a session where the source document includes a client’s account number, and have the rubric check whether the learner redacted it before prompting. Employees get that decision wrong in practice, get coached, and get it right before it matters.

Training should own the employee’s side of the line. Questions about how Claude processes inputs belong with your vendor contract and IT security team.

Set up Claude training in Yoodli

1. Define one concrete outcome per session. “Learn Claude” is too broad. “Summarize a public 10-K in Claude and verify three figures against the filing” works.
2. Open the Roleplay Builder and choose the Tutor template.
3. Write the lesson plan in the context field. It should cover the greeting, the order of the interface walkthrough, when to switch to the learner’s screen, the hands-on task, and how to coach during practice.
4. Upload screenshots of the Claude screens the tutor will reference.
5. Set the tutor’s name and tone.
6. Add rubric criteria to certify learners.
7. Add the sessions to a Program to sequence them and track completion.
8. Run the full session yourself before launch.

The Yoodli help center guide to platform certification walks through each step. The Yoodli template library includes tool training templates to start from.

What results look like

Yoodli customers use the same certification approach across large rollouts. Google Cloud certified more than 15,000 employees on a new pitch in under a month. RingCentral cut call center certification time by 90%. Harness reduced training review time by 75%.

Faster certification shortens time to full productivity, and that matters more in financial services. Every week without practice is a week of employees making AI judgment calls without the reps to back them up.

Expand as the approved list grows

After a few months, compliance will have visibility into how employees actually use Claude and will start approving new use cases. Each one follows the same path. Compliance reviews it, you build a Yoodli session around it, and employees certify on it. If a session surfaced confusion or a near miss on data handling, that goes into the next version of the rubric.

Regulatory guidance on generative AI keeps changing, so track it at the source. FINRA maintains artificial intelligence guidance for member firms, and it’s worth checking whenever your use case list expands.

Frequently asked questions

How do you train employees on Claude? The most effective way is hands-on practice inside Claude with coaching as employees work. Yoodli’s AI tutor walks learners through the interface, assigns tasks based on your approved use cases, watches over screen share, and certifies them against your rubric.

Can Yoodli certify employees on Claude? Yes. Learners complete real tasks in Claude while Yoodli evaluates them against criteria your team defines, and Programs track who has certified.

Does Yoodli train on tools other than Claude? Yes. The same setup works for ChatGPT, Copilot, Gemini, and internal tools.

See how Yoodli trains and certifies teams on Claude and other AI platforms, or talk to our team about your rollout.