Hiring an AI Learning Experience Designer Who Changes How People Actually Work

Why does your AI training keep getting perfect attendance and changing nothing?

Because the training was written for nobody in particular. Ninety percent completion on a two-hour generative AI course, glowing survey scores, and six weeks later the underwriting team is still rekeying the same figures by hand. An AI Learning Experience Designer exists to close that gap, and the gap is the entire justification for the headcount.

Here is the moment that usually triggers the requisition. A VP has bought seats on a general AI literacy platform, run everyone through it, and now has to answer a board question about return. The completion dashboard is green. The process metrics are flat. Nobody in the room can say which specific task anyone does differently, because the course never named a task: it taught prompt patterns in the abstract to an audience of claims adjusters, schedulers, recruiters and field techs at once.

Two of the numbers everyone quotes here are worth carrying and one is worth arguing with. The World Economic Forum expects around 40% of the skills a worker needs to change by 2030, with 77% of surveyed employers planning to reskill their people to work alongside AI ¹, and that is the size of the problem landing on a function which has spent a decade optimizing for completions. Lightcast's 200% growth figure for generative AI roles inside education and training ² is a hiring signal rather than a capability one, and part of what it describes is the same undifferentiated course arriving faster.

So the job description you write matters more than the title you pick. Ask for curricula built one occupation at a time, a mobility path attached to each one, and a stated measure that is not attendance. The rest of this article is what to look for in the person who can deliver that.

What Does a Portfolio Review Show That an Interview Cannot?

A portfolio review answers this in about twenty minutes if you know what to look for. The designer worth hiring builds from a task inventory rather than a tool list, states a behavior change as the success measure before the program launches, and can walk you through a rollout that failed and what changed afterward. The one to pass on leads with tools, platforms and completion charts.

The task inventory comes first. Ask any candidate how they would build AI training for your accounts payable team, and listen for the first question back. The useful one is about the work: what the team does all day, which parts are judgment and which are transcription, where errors currently get caught. Naming models and platforms first is the answer of someone who will write the course they wrote last time with a new cover on it. That instinct is the whole job in miniature, because the curriculum is a claim about which tasks change, and you cannot make that claim without watching the work.

Then push on the success measure. "Learner confidence rose" is a survey artifact. "Cycle time on renewals fell and four schedulers moved into the exceptions queue" is a claim someone can check against a system of record. The designers worth hiring name the measure before the build, agree it with the business owner, and accept that it might come back flat.

The failed rollout is the single most reliable filter in the whole loop. Ask for a program that did not work and what specifically was wrong with it. Real practitioners answer immediately and with unflattering detail: the pilot cohort was volunteers so the results did not generalize, or the AI assistant in the exercise was too well-behaved to teach anyone what overreach looks like. A candidate who has only shipped successes has either not shipped much or is not going to tell you the truth about the next one.

Quieter, and worth asking last: what does the training say about when not to use the assistant? Anyone who has actually taught this has a section on it, because the first month of any rollout produces at least one person pasting something they should not have into a chat window, and at least one confidently wrong output that shipped. A curriculum with no restraint in it was written from a vendor deck.

Which backgrounds produce this designer, including the ones you would not post for?

Instructional design and corporate L&D are the obvious feeder, and they supply most of the market. The less obvious feeders are better than their resumes suggest: frontline operations leads, technical support and enablement, community college and workforce-development instructors, and internal comms people who have run a change program end to end.

The instructional designer brings the craft that is hardest to fake. Task analysis, assessment design, scaffolding, knowing that a demonstration is not practice. If they have added AI fluency on top of that foundation, they are the safe hire. eLearning Industry's 2026 read on the field describes exactly this shift, with instructional design demand holding and AI-inflected variants of the title expanding, and hiring moving toward portfolios of real projects with measurable outcomes rather than credentials ³.

The unexpected candidates earn a look for a specific reason. A former operations lead in claims or logistics knows the work in a way no curriculum designer can learn in a week, and that domain knowledge is the scarce half. Teaching craft can be coached; credible standing with a skeptical floor team cannot. Same logic for a support or enablement veteran, who has spent years watching where people actually get stuck and has probably built more usable job aids than most L&D departments.

A note on the community college and workforce-development instructor. That person has taught skills to adults who did not choose to be in the room, on a budget, with outcomes measured in placements. That is your reskilling program's exact shape. They will need help with corporate stakeholder management, and they are frequently underpriced relative to what they can do.

Who to be careful with: the pure prompt-engineering trainer with no assessment background, and the platform administrator whose experience is configuring an LMS. Both can be genuinely useful on a team. Neither has usually had to prove that behavior changed. If you are also standing up measurement across hiring and internal mobility, this role pairs naturally with an AI skills assessment specialist, and the two jobs are often confused in job postings written by people who need both.

How did this person get good, and what does their own AI practice look like?

They used AI to compress the slowest parts of their own craft, then kept the parts where their judgment was the product. Concretely: drafting scenario variants at volume, generating distractors for knowledge checks, personalizing practice to a role, and building a simulated assistant that behaves badly on purpose so learners get practice catching it.

Ask them to describe the last thing they built with an assistant, in detail, and the answer tells you almost everything. The practiced version sounds like this. They wrote twelve scenario stems by hand for an insurance underwriting course, used a model to expand each into three difficulty variants, then threw out roughly half because the model produced plausible-sounding cases that could not actually happen under that state's filing rules. The throwing-out is the skill. Anyone can generate forty scenarios; knowing which ones a subject matter expert will laugh at is what takes practice.

The habit worth probing next is the deliberately unreliable assistant. Good AI training does not hand learners a well-behaved model, because a well-behaved model teaches nothing about verification. Designers who have taught this build exercises where the assistant cites a regulation that does not exist, or produces a confident number with no traceable source, and the learner's task is to catch it. If a candidate has built that, they will be eager to tell you about it. This same instinct shows up in the engineering roles that run assistants in production, where the discipline of catching an overconfident output is a job requirement rather than a lesson; an AgentOps engineer is a useful partner for a designer building those exercises against a real system.

Personalization needs a boundary, and the boundary is the interesting part. Personalizing practice by role and by current proficiency puts people in front of work they cannot yet do. Personalizing by learner preference mostly reproduces what they already do well, which feels good and teaches nothing. Ask which one they built and why.

What none of them should claim: that a course or a tool can tell you whether a document was written by a model. That capability does not work reliably, and a candidate who sells it as a training outcome has a credibility problem you will inherit.

Where do you source an AI Learning Experience Designer, and what closes the offer?

Source from practitioner communities rather than job boards: the Learning Guild and ATD conference circuits, the LX Design and instructional design communities on LinkedIn, university learning-design teams, and your own frontline. Close on scope and access, because the thing this person actually wants is a business owner who will let them near the work.

The named venues are worth the travel. Learning Guild's DevLearn and ATD's annual conference are where portfolio-carrying practitioners congregate, and the sessions themselves are a screen: whoever is presenting a measured case study rather than a tool demo is who you want to talk to. LinkedIn is the tempting shortcut and the weakest of the venues, precisely because it is working: one January 2026 report puts engagement with AI learning material on the platform up 92% year over year ⁴. That number measures interest, and interest is the one thing you cannot screen on. Expect a large self-identifying pool and a low hit rate.

The higher-yield source is internal. Someone in your operations org has already built the unofficial job aid that everyone actually uses, and they know your work in a way an external hire will spend six months acquiring. Post the role internally first and genuinely mean it.

On closing. What kills these offers, in rough order of frequency: a reporting line that buries the role three levels down in HR with no route to the business owners, a mandate that is really "roll out the platform we bought," and a success metric of completions written into the objectives. Any of those and your best candidate withdraws, usually politely, usually citing something else.

What closes them is small and specific. Name the first occupation they will redesign for and the business leader who has agreed to sponsor it. Commit to a measure that is not attendance and put it in writing. Give them budget for subject matter expert time, which is the resource that actually constrains this work. Offer a path: many strong candidates take this job because reskilling design is a route toward workforce strategy, and saying so plainly costs you nothing. If your organization is in a regulated hiring context, they will also want to know who they can ask when a training decision touches employment law, which is where an AI hiring compliance manager earns their seat.

What should you pay an AI Learning Experience Designer, and does the role sit remote?

No public wage series exists for this exact title yet, so every number below is a proxy: price against instructional design and learning experience design bands and add for AI fluency. One vendor benchmark, published by the GSD Council off roughly 11,000 self-reported Glassdoor records as of June 2026, puts average instructional designer pay near $92,670, with the middle of the market between about $74,000 and $117,000 ⁵.

Read those as a floor rather than a target, and remember they price the instructional designer title rather than this one. The same benchmark places senior instructional designers roughly in the $95,000 to $120,000 range and lead or LXD-level roles at about $115,000 to $145,000, with L&D director roles above that ⁵. It also reports AI and generative AI fluency in the workflow correlating with a $15,000 to $25,000 uplift independent of certification or seniority ⁵. Aggregator data carries the usual caveats: self-reported, US-weighted, and blind to your geography. Treat it as a starting band, confirm against two other sources for your market, and expect a designer with genuine domain depth in a regulated industry to price above it.

One broader signal, dated mid-2025 and worth carrying with its hedge: Lightcast found job postings that include AI skills advertising about 28% higher salaries, close to $18,000 more per year, than comparable postings without them ². That is across postings generally rather than this title specifically, so use it to argue the shape of the premium rather than the number.

On location. The design and build work is remote-friendly and most of these roles are posted that way. The observation work is not. Task inventories, subject matter expert sessions and pilot facilitation go far better in the room, particularly with frontline teams who have watched initiatives arrive from headquarters before. The practical norm settling in for 2026 looks like remote-default with structured onsite blocks: a week with each occupation at the start of a build, and presence for the first pilot cohort. Write that into the posting. A candidate who reads "fully remote" and then discovers monthly travel to three plants will not stay, and a candidate who would have loved the travel never applied.

See a sample report >