AI simulation authoring for learning teams

Turn your team's knowledge into realistic job practice.

Build voice and text simulations from your SOPs, call recordings and courses. Score each attempt against your criteria and certify people at scale without a trainer in every room.

Tell devlin the result you need. It designs the practice and explains why.

Book a demoBuild a simulation
  • SOC 2 Type 2
  • SAML SSO
  • Scores to your LMS
  • 2,500+ instructional designers

Used by L&D teams in retail, healthcare, government, and education

From a request to measurable impact

A walkthrough of devlin.ai in four chapters, shown with sample data.

  1. Ask devlin: A designer tells devlin that new store associates keep mishandling returns without a receipt and need to be certified before the holidays, and drags in the returns policy PDF, the Rise course as a SCORM package, and a recorded call with a subject matter expert. devlin recommends a voice role-play with an upset customer, lists the three things it will score, and explains why the design works. The designer clicks Build now and the draft builds, with its progress shown step by step.
  2. Practice by voice: A learner reads the scenario and starts a voice session with Maria, a customer who lost her receipt. Maria speaks, the learner answers, and the transcript fills in as they talk. Beside the call: what the learner will be scored on, which is verifying the purchase, explaining the policy and offering an alternative.
  3. Score and coach: The attempt is scored one criterion at a time, with points and a one-line reason: 30 of 30, 30 of 30 and 20 of 40. The learner passed with 80 and gets a coaching note.
  4. Measurable results: The results add up at three levels. Learning data shows completions, pass rate and average score, with full transcripts and a per-criterion breakdown. Skill-level insights show the skill at 84% against a target of 80%, with a score for each sub-skill. Business metrics show returns escalated to a manager falling from 14% to 10.2% against a target of 9%, and refunds within policy rising from 78% to 91%.

An illustration of one simulation, from request to results, with sample data.

Why traditional practice fails

Role-plays don't scale

Every observed role-play costs a trainer's hour. Twice the learners means twice the hours.

Quizzes scale but measure recall

A learner can pick the right answer and still freeze in the real conversation.

Completion rates are all you have

When the VP asks who can actually do the job, a completion rate doesn't answer it.

See AI practice in action

Describe a conversation your learners need to practice, then try the simulation right here.

What you can put in front of your VP

Track business metrics, skill proficiency and learning over time.

Returns escalated to a manager

0%5%10%15%JunJulAugSepOctBaseline 14%Target 9%10.2%

Practice score

61% to 84%

Refunds within policy

78% to 91%

Sample data. You record business numbers from your own reports.

Practice for the conversations the job depends on

Certification and recert

Learners show they can follow the procedure in conversation and are scored against the same criteria every time.

New-hire ramp

New hires practice the conversations of the role before their first live customer.

Patient and client conversations

Practice intake, bad news and upset clients by voice or text until they're routine.

Manager conversations

Practice feedback, performance reviews and conflict before the meeting that counts.

See every use case

Works wherever you train

In your course, in your LMS, on a phone, or in the room. Results come back to you either way.

Storyline and Rise

Embed in your course. Results go to Storyline variables or your LMS.

Your LMS

SCORM packages report completion and score.

Phones

Learners practice by text or voice on any device, with no app to install.

Live sessions

Share a link and the whole room practices during class. Review results after.

Hosted links

Send a standalone link in email, Slack, or Teams.

Your AI

Build and manage simulations from Claude or ChatGPT.

Questions your team will ask

Will it pass our security review?

devlin.ai has a SOC 2 Type 2 report, available under NDA in our Trust Center. Enterprise teams sign in through their identity provider with SAML SSO. Customer data is stored in the US. See our security page.

How do we know the scores are right?

Characters answer from the documents you attach. Every score shows the reasoning for each criterion next to the full transcript. Owners and admins can review a result and override it with a reason. If you're not ready to score, feedback-only mode gives written feedback with no score.

We already use Articulate. Does this replace it?

No. devlin.ai works inside your Articulate courses. In Storyline, results are written into your course variables. In Rise, results reach your LMS through the SCORM export.

Why not just use ChatGPT or Copilot?

A general chat assistant can play a character, but it has no rubric, no score in your LMS, and no transcript per learner for you to review. devlin.ai gives you all three. If your team already works in Claude or ChatGPT, you can build devlin.ai simulations from there through our MCP connector.

Scope a pilot on one team.

Bring one role and one conversation that matters. We'll scope the pilot with you on the call.

Book a demo