Skip to content

Enablement and training for builders

Seat licences alone underdeliver. In surveys, regular use goes with meaningful training plus coaching, and manager support is associated with more use than tool access alone (as of October 2026). Run short role-based tracks for builders, app owners, business reviewers, technical reviewers and managers. Fund champions with protected time, hold weekly office hours and make managers sponsors. The training hours on this page are hypotheses to test in your pilot, not proven figures.

Who this is for

Enablement leads, IT managers, managers of builders, and HR and learning partners building a citizen developer training program. You get five training tracks, how to run champions, a plan for office hours and community, the manager's role and how to measure enablement. Reading time: about 9 minutes.

What does the evidence say?

  • Licences without enablement fall short. A UK government evaluation of 1,000 licences found time saved on some tasks but no clear evidence of productivity gains.1 Only 30% of participants used the tool at least once a day.3 The evaluation doesn't say why. Our reading, not the report's, is that licences without training or task redesign buy satisfaction rather than productivity. A cross-industry survey links five or more hours of training plus coaching to regular use.2
  • Several forms of training help. In an Australian government trial, 75% of participants who received three or more forms of training said they felt confident (non-randomized; self-reported). We couldn't fetch the source page on 2026-10-22; the figure is taken from the search-index text of the evaluation's executive summary, so treat it as unverified.4
  • Managers matter. Gallup finds employees whose managers actively support AI are 1.7 times as likely to use it a few times a week or more, and BCG reports positive attitudes rising from 15% to 55% where leaders back it. Both are surveys, not experiments.52
  • Communities can grow with little central cost reported. A UK cross-government low-code community reached about 700 members through weekly sessions and shared repositories, largely on volunteer effort, with no outcome data on policy compliance.6

Most of this is survey and government self-report. It comes from AI assistants and low-code, not from non-developer builders of AI-built apps. That's why the pilot measures local results.

What training tracks should builders take?

Keep each track short and tied to a real task. Hours are suggested starting points, a hypothesis to test (A-023). The evidence supports "more than a launch session, with coaching", not a particular number.

Hypothesis to test

We expect about five hours of training plus coaching to be enough for a builder to use AI regularly, and less for the other roles. Test this in your pilot with staggered training waves: compare sustained use and intake quality for builders trained before building with those trained at day 45.

Track Who Outcomes Suggested hours (hypothesis) Pages it covers
Builder Anyone who builds or changes an app, non-developers and developers Knows the five never-dos, what needs a yes, how to answer intake, write a spec, test, document and run the self-check About 5, plus coaching on the builder's own first app Build an app with AI at work, intake, verify
App owner People accountable for an app in use Owner duties, a named successor, the support level their tier gives About 1 App ownership, support model
Business reviewer Colleagues who review logic and data use at Important and Critical What to check, what a review must show, when to stop and escalate About 1.5 Business review and technical check
Technical reviewer Platform engineers and champions on the published list Does the Important-tier technical check: an exposure scan, data-rule negative tests, the fixed-list check About 4, then shadowing two reviews. An IT technical reviewer co-signs the champion's first three checks and re-checks a sample after that Risk tiers, roles
Manager Managers of builders Sponsor's duties: protect time, approve intake, name an owner when someone leaves About 1 Roles and decision rights

The technical-reviewer track is what keeps Important-tier review capacity from becoming the bottleneck (A-042). Train enough champions that the median wait for a technical check stays within days, and publish the list.

Don't: treat one launch session as training

A single kickoff with no follow-up gives builders the least to go on: in the Australian trial, confidence rose with the number of forms of training people received.4 Combine a short course, practice on the builder's own task, and office hours.

Don't: train on the tool and skip the rules

Product mechanics are the easy part. The rules that protect the business (never-dos, data classes, tiers) belong in your own session, taught by your own people.

How should champions work?

A champion is a builder with extra training who helps others. Champions give team-supported help at Team tier, and trained champions on the published list can act as technical reviewers.

  • Selection. Ask managers to nominate, then confirm the person wants it. Pick across business units so no team lacks one. Avoid choosing only the most enthusiastic builders.
  • Protected time. Agree it with the champion's manager in writing, and record it. Start from the hypothesis below and adjust after the pilot.
  • What they do. Run office hours in rotation, review Team-tier apps informally, coach first-time builders, and do technical checks once trained and listed.
  • Recognition. Name them in team communications, note the work in performance conversations and give them a route to the enablement lead.
  • Succession. Name a deputy for each champion, so one departure doesn't empty a unit.

Hypothesis to test

We expect a champion to need roughly 2 to 4 hours a week: one office-hours slot, some coaching and, once listed, a technical check or two. This is our starting estimate, not a sourced figure. Test it in your pilot by logging champion hours each week, and record the result with the pilot's cost figures.

Don't: run volunteer-only champions with no time

We expect champions with no protected time to stop when their own work gets busy (A-024, a hypothesis). The evidence is thin both ways. One department reports sponsoring a community lead and setting aside a budget for its maker community.7 The UK cross-government community grew on volunteer effort, and it reports no outcome data.6 Test the funded model in your pilot rather than assuming it.

How should office hours and community work?

  • Weekly office hours, open to anyone building, run by champions in rotation, with a standing agenda: new tools, a reviewed example, open questions.
  • A shared catalog of approved examples and starter prompts, so builders reuse instead of duplicating. Link to the starter prompts on Write a build spec rather than writing your own copy.
  • A structured start for new builders. One vendor's onboarding guide describes a six-week pattern: welcome, getting-started guide, Q&A, mentor pairing and recognition. It's cheap to copy, but its effect on app quality is asserted, not measured.8
  • A visible channel for questions and for reporting apps nobody owns.

What is the manager's role?

Managers sponsor. They agree the time to build, make sure the builder's tier and owner are recorded, approve intake for their staff's apps, release champions for their hours, and name a new owner when someone leaves. Give them the manager track, and ask them for one thing per quarter: confirm their team's registered apps still have owners. Surveys connect this support to use more than tool access does (A-022).

How do I measure enablement?

Measure outcomes, not attendance.

  • Sustained use by trained and not-yet-trained staff, from tool logs, over at least a quarter.
  • Intake quality: the share of intakes that need rework or a corrected tier.
  • Review wait: median days for a business review and for a technical check.
  • Orphans: apps with no current owner after a quarter.

Don't: use attendance or course satisfaction as the success measure

A UK evaluation found satisfaction without a clear productivity gain.1 See baselining for how to compare against a starting point.

Depends on your circumstances

Depends on your circumstances (C7)

If several business units already build independently, run one shared community and one set of tracks, with a champion in each unit. Units may add local material, but the never-dos, data classes and tier questions stay identical everywhere.

Depends on your circumstances (C9)

If several AI tools are already in use, teach the common rules once and add a short tool-specific session for each approved tool. Keep tool mechanics out of the core track so it doesn't date when a tool changes.

Does the EU AI Act require this training?

Possibly. Article 4, as amended by the Digital Omnibus, asks for measures to support AI literacy where it applies to you (as of October 2026), and these tracks can be part of the evidence. Whether it reaches you, and what is enough, are questions for your counsel. See deployer duties under the EU AI Act for the details.

Record the decision

Illustrative composite

Title: Training and champion commitments · Status: proposed · Date: YYYY-MM-DD · Owner: enablement lead (a role)

Context. Staff build apps with AI, and review capacity and consistency depend on trained people.

Decision. Adopt five role-based tracks, champions with written protected time, weekly office hours and a published technical-reviewer list. Treat the hours as hypotheses and adjust after the pilot.

Alternatives considered. Seat licences only (rejected: weak use evidence). One launch session (rejected: no follow-up). Vendor-only training (rejected: doesn't teach our rules).

Consequences. Managers release time, and the enablement lead keeps the list current. The pilot trains builders in two waves, and we compare sustained use between them at day 45.

Review by: six months from the date above.

Assumptions and evidence

  • A-022 (Medium): we expect manager support to beat tool access as an adoption lever. Test this in your pilot by training half the managers at the start and the rest at day 45, and comparing their teams' sustained use at day 45.
  • A-023 (Medium): we expect training to be needed for regular use. Test this by comparing builders trained before building with builders trained at day 45.
  • A-024 (Low): we expect a named, funded community with champions to raise paved-road compliance more than mandates. A pilot can only observe this, because units outside it also lack the paved road. Test it at the next stage by phasing champion funding in unit by unit.
  • A-042 (Low): we expect Important-tier review capacity to keep up. Test this by tracking the median wait for a technical check against five working days.

Evidence: government evaluations and surveys, some vendor-run; no controlled study of non-developer builders. Facts on this page are as of October 2026.

  1. Civil Service World, "DBT trial finds Copilot saves staff time but doesn't improve productivity" (2025-09). https://www.civilserviceworld.com/news/article/dbt-trial-finds-copilot-saves-staff-time-but-doesnt-improve-productivity [news], accessed 2026-10-22. ↩↩

  2. BCG, "AI at Work: Momentum builds, but gaps remain" (2025). https://www.bcg.com/publications/2025/ai-at-work-momentum-builds-but-gaps-remain [survey], accessed 2026-10-22. ↩↩

  3. The Register, "UK government trial of M365 Copilot finds no clear productivity boost" (2025-09-04): "30 percent used it at least once a day". https://www.theregister.com/2025/09/04/m365_copilot_uk_government [news], accessed 2026-10-22. ↩

  4. Australian Digital Transformation Agency, Microsoft 365 Copilot trial evaluation, executive summary (report Oct 2024; non-randomized, not coding-specific). The page timed out on 2026-10-22; wording checked only against its search-index text. https://www.digital.gov.au/initiatives/copilot-trial/microsoft-365-copilot-evaluation-report-full/executive-summary-glossary [government], accessed 2026-10-22 (unverified). ↩↩

  5. Gallup, AI use at work indicator (data as of May 2026): employees whose managers actively support AI are "1.7x as likely to use AI a few times a week or more". https://www.gallup.com/699797/indicator-artificial-intelligence.aspx [survey], accessed 2026-10-22. ↩

  6. UK cross-government low-code community, "Celebrating one year of the cross-government low-code community" (2026-02-06). https://technology.blog.gov.uk/2026/02/06/celebrating-one-year-of-the-cross-government-low-code-community-a-people-first-approach [primary], accessed 2026-10-22. ↩↩

  7. Government Transformation, "Findings from the cross-government low-code workshop" (2025-11-04): reports a maker community of 1,000+ at one department; self-reported; one author is a vendor employee. https://www.government-transformation.com/transformation/findings-from-the-cross-government-low-code-workshop-opportunities-risks-myths-and-success-stories [news], accessed 2026-10-22. ↩

  8. Microsoft, "Onboard makers" (2025-05-14), including champions, mentors and office hours. https://learn.microsoft.com/en-us/power-platform/guidance/adoption/onboard-makers [vendor], accessed 2026-10-22. ↩