Training and Development · the service
We address the condition that limits each team, and we measure whether that condition moved.
The condition that limits each business area (the binding constraint) is named by the profile. Training and Development is our work on that condition, team by team – twenty modules in four packages, of which training is one.
What the service covers
Each of the four conditions is measured by AI-MARS through five sub-conditions, and each sub-condition has one module (twenty modules in four packages). A team receives the package for its binding constraint, and within that package the modules its low sub-conditions call for. The other three conditions in that team are left alone.
No remedy is delivered before the first census has been taken – the remedy is chosen from the profile. Thus, the account and the census come first. The account costs nothing, and the census uses founding places until the 3,000 are taken.
From the profile to the remedy
In one organisation, four teams need four different modules.
The profile below is illustrative (an organisation of 60 people, four business areas, 51 responses).
Example Organisation · first census51 of 60 responded · illustrative data
- Customer service (n = 19)Role perceptions bind with a score of 31; lowest sub-condition: policy clarity (score 18)Policy launch and the safe-use session
- Finance (n = 9)Ability binds with a score of 29; lowest sub-condition: use self-efficacy (score 21)Critical-task training
- Contracting (n = 12)Motivation binds with a score of 38; lowest sub-condition: social influence (score 24)Champions
- Sales (n = 11)Situational support binds with a score of 33; lowest sub-condition: time availability (score 19)Released-time plan
Of these four teams, one needs training. A single course for all four teams would miss the binding constraint in the other three.
How the remedy is chosen
The binding constraint is the lowest-scoring condition in the business area (the red bar). Each bar is the area's score for that condition, from 0 to 100, read against the target level, and every condition below the target limits the area. The four conditions are never added together, due to the fact that a strong average hides the condition that is absent.
The sub-condition scores beneath a condition show which part of that condition is low, and each sub-condition has one module that answers it (twenty modules in all, five per condition). A low motivation score driven by strain calls for a different module from a low motivation score driven by what peers do. Where two conditions in one area are within a few points of each other, the two packages are sequenced.
Twenty modules in four packages
Each sub-condition of AI-MARS has one module that answers it, and each condition has one package of five.
A team receives the package for its binding constraint, and within the package the modules its low sub-conditions call for.
Motivation · Do I want to?
The motivation package
Where motivation binds, we arrange early experiences of usefulness on the person's own tasks, and champions drawn from the same peer group describe their own use of the tools. What peers did was the largest predictor of intention (β = 0.66) in a survey of 300 employees of Korean companies. What peers did was also the only factor tested that was associated with actual use (Kim et al., 2024).
| Sub-condition that is low | What it looks like in the team | The module | When |
|---|---|---|---|
| Performance expectancy | Staff cannot see where a tool would help their own work | Usefulness demonstrations: a session per team working three of its own critical tasks with the tool, timed before and after | Embed, weeks 8–20 |
| Social influence | No colleague they respect uses it yet | Champions: one per area, chosen on standing, coached fortnightly for three months, with a briefing so managers use the tools where staff can see them | Embed, weeks 8–20 |
| Output trust | The output is not trusted enough to bother | Calibration clinic: the tool tested on tasks with known answers, and a team note on where the tool is reliable | Train, weeks 6–14 |
| Technostress | Worry about data, jobs or being judged | Assurance on data and jobs: the data-handling rules and the board's statement on jobs, delivered to the team with anonymous questions | Rules, weeks 1–4 |
| Exploratory use | Willing, but nothing has been tried | Trial slots: protected time to try one task, with a log, run through the champions | Embed, weeks 8–20 |
Read more
The association between what peers did and intention was found in the same study to be stronger among staff over 40 and among staff with more than five years' experience. This suggests that the longer-serving staff are the ones most likely to move when a peer moves first. As such, champions are chosen for their standing among their peers, and one champion is selected for each business area to begin with.
Our usefulness demonstrations will use the team's own critical tasks (drawn from the profile and from a short session with the team), and the first tasks are chosen so that usefulness shows early. The motivation package will ask about two hours a fortnight from each champion for three months – and nothing from anyone else beyond their normal work.
Ability · Can I?
The ability package
Where ability binds, each cohort is trained on its own critical tasks, in two sessions with structured practice on real work between the two sessions. Knowledge was named as the primary barrier by every participant in a study of Malaysian SMEs (small and medium-sized enterprises; Falahat et al., 2026). Self-assessed AI knowledge was the largest predictor of intention in a study of 488 teachers (β = 0.79; Kilat, 2026).
| Sub-condition that is low | What it looks like in the team | The module | When |
|---|---|---|---|
| AI literacy | A wrong picture of what the tools can do | Foundations by role: what the tools are good at and bad at, on the team's own kinds of task | Train, weeks 6–14 |
| Use self-efficacy | Requests produce nothing usable | Critical-task training: two sessions per cohort with structured practice on real work between them | Train, weeks 6–14 |
| Output verification | Errors pass unnoticed | Verification training: a checking protocol per task type, error-spotting exercises and a team checklist | Train, weeks 6–14 |
| Learning self-efficacy | People stall and do not ask | Help route: a go-to person, office hours and pair practice | Train, weeks 6–14 |
| Practised use | Trained, but not practising | Practice log: real work between the sessions, logged and reviewed in coaching | Embed, weeks 8–20 |
Read more
Each cohort receives a different course, due to the fact that the critical tasks differ from one role to the next and the low sub-condition within ability differs too (AI literacy, use self-efficacy, output verification, or learning self-efficacy). The sessions are sequenced with confidence first and usefulness second. This order is suggested for staff at the early stage of adoption by the evidence on ease of use.
For an organisation of 60 people, the role-specific training will run as six cohorts of about nine people, with two sessions per cohort and structured practice on real work between the two sessions.
Role perceptions · Is it my job, and am I allowed?
The role perceptions package
Where role perceptions bind, the staff do not know whether the work is theirs or whether the work is allowed. Use of a tool on a personal account usually means that the rules were unclear. The organisation's policy is written from what our audit of the tools in use finds, the rules are restated for each role, and the approval loop is put into the workflow, with sign-off by a named person.
| Sub-condition that is low | What it looks like in the team | The module | When |
|---|---|---|---|
| Policy clarity | Nobody knows whether it is permitted | Policy launch and the safe-use session (60–90 minutes, with a knowledge check and a dated record of attendance) | Rules, weeks 1–4 |
| Task-scope clarity | Allowed in general, unclear in particular | Task register: for each role, the tasks that are allowed, allowed with approval, or never, drawn from the registry of tools crossed with the four data classes | Rules, weeks 1–4 |
| Role expectation | Seen as someone else's work | Role objectives: role statements and objectives updated, and the board's statement on what AI means for jobs | Rules, weeks 1–4 |
| Accountability clarity | No named person to ask | Approval loop: in the workflow, with sign-off by a named person | Rules, weeks 1–4 |
| Compliant use | Rules known, not followed | Repeat audit: at three months, under the same amnesty as the first | Embed, weeks 8–20 |
Read more
A list of approved tools is asked for in most briefs. A list drawn up before our audit can draw only on the subscription record, however the subscription record shows paid accounts and not use on personal accounts. Thus, our audit comes first (a census, under a signed amnesty announced before the questionnaire is issued), and the approved-tools list is written from what our audit finds.
The role perceptions package will also produce an accreditation map. Our accreditation map lists every certificate the organisation has (Cyber Essentials, or sector schemes such as JOSCAR, CHAS and FIS), what each scheme requires of AI and cloud use, and the gaps between those requirements and present practice. The dated record of attendance at the safe-use session is kept as evidence of dissemination for the accreditation assessments.
Governance is itself an adoption intervention. Unanswered questions about jobs are likely to reduce adoption, however reduced adoption of this kind is easily mistaken for a training problem. These questions are answered by the board's statement on jobs (which forms part of the policy). Where the board's statement names where the freed hours will go, the staff have a claim they can check.
Situational support · Do I have what I need?
The situational support package
Where situational support binds, the people want to use the tools, know how to and are allowed to, however the licence, the time or the cover is missing. The situational support package is the least costly of the four packages and the one most often skipped. The package consists of the licences, time released by the line manager (sized from two weeks of work sampling) and a named sponsor.
| Sub-condition that is low | What it looks like in the team | The module | When |
|---|---|---|---|
| Tool access | No work account, or the tool does not fit the task | Tool-fit audit: licences, and where a tool does not fit a task, an approved route for that task with an approval step that cannot be bypassed | Enable, weeks 3–8 |
| Time availability | No slack in the week | Released-time plan: two weeks of work sampling, and released time agreed with each person, through the line manager, from what the sampling shows | Enable, weeks 3–8 |
| Leadership support | Words from the top, no cover | Sponsor briefing: a named sponsor for each business area, and one session with the leadership team | Enable, weeks 3–8 |
| Psychological safety | Failed trials are hidden | Failed-trial review: line-manager practice on failed trials, and a blameless review of each trial | Enable, weeks 3–8 |
| Enacted support | Promised time never arrives | Provision check: monthly, against what was promised, and checked by the sponsor | Embed, weeks 8–20 |
Read more
Psychological safety is measured under this condition. AI-MARS asks whether a person could say that a trial went badly without that admission counting against them. Where that sub-condition is the low one, the remedy is a named sponsor and a change in how the line manager responds to a failed trial, not a licence.
When each module is delivered
A programme delivers the same twenty modules in four overlapping phases.
The rules come first for everyone, and each team is then enabled, trained and embedded in turn. The phases overlap by a few weeks, and a team receives only the modules the team's profile calls for.
Rules
Everyone, once. The policy and the rules exist before any team is asked to use the tools.
Assurance on data and jobs
Policy launch and the safe-use session- Task register
- Role objectives
- Approval loop
Enable
By team. The licence, the time and the cover are put in place before the training, which would otherwise fail for want of them.
Tool-fit audit- Released-time plan
- Sponsor briefing
- Failed-trial review
Train
By cohort, in a randomised order. The cohorts are sorted by role and by the low sub-condition in the profile.
- Calibration clinic
Foundations by role- Critical-task training
- Verification training
- Help route
Embed
By team, and carried on by the champions after we have gone. Each team is measured at four and twelve weeks.
- Usefulness demonstrations
- Champions
- Trial slots
- Practice log
- Repeat audit
- Provision check
How a programme runs
The order in which teams start is drawn by lot, and the teams not yet started are the comparison.
A programme runs in five steps, and no remedy is delivered before the baseline has been taken.
-
The first census
The organisation will run AI-MARS through the portal. The organisation's profile will name the binding constraint in each business area, and the baseline will be agreed.
-
Remedies chosen by team
Each team will be matched to the remedy for its binding constraint, and the indicators for each team will be fixed in writing before anything is delivered.
-
Delivery in a randomised order
The teams will be treated one after another in a randomised order. The teams not yet treated are the comparison for the teams already treated.
-
Measurement at four and twelve weeks
Behaviour and artefacts (the quotes produced, the checks run, the process followed) will be measured against the indicators fixed at step two.
-
The second census
A second round of AI-MARS will show which conditions moved in which teams, and the comparison set will show where the organisation now stands.
Why the order is randomised
Training follows a stepped-wedge design (Hussey and Hughes, 2007). The business areas are treated in a randomised order, and everyone is treated by the end. The areas not yet treated are the comparison for the areas already treated. Thus, each business area contributes observations before and after its own crossover. These observations allow a change to be attributed to the remedy and not to the season, the order book or a new manager.
It should be said that the timing shown is illustrative. The timing for a given organisation is set with the organisation around its leave and busy periods, and where the profile shows that two teams share a binding constraint, the two teams may be treated as one cohort.
The measurement rule
Satisfaction with a session is not evidence that the work changed.
Affective reactions to training correlate with transfer at about 0.07 (Alliger et al., 1997). Thus, satisfaction is used for quality control, and our measurement is of the work.
What is measured instead
No remedy is delivered before the baseline is taken. Measures are then taken at four and twelve weeks, in behaviour and in artefacts (the quotes produced, the checks run, the process followed). The indicators are fixed in writing before the remedy starts – the result cannot then be chosen after the fact. Where the indicators have not moved at twelve weeks, our report will say so and will give the likely reason.
- 0remedies delivered before the baseline has been taken
- 4 wkfirst measurement, in behaviour and in artefacts
- 12 wksecond measurement, against the same indicators
- 0.07the correlation between satisfaction with a session and transfer to the work
What it costs
The census uses founding places until the 3,000 are taken, and a programme is scoped from the profile and quoted per team.
The first 3,000 employees measured through the portal, across all organisations, are not charged, and once those founding places are taken, the census is priced. A programme is priced once the profile shows which teams need which remedy, due to the fact that the price depends on how many teams need training and how many need only the rules restated. It is stated before each step what the step will produce, what the step will ask of the business and how we will know whether it worked. The organisation decides at the end of each step whether to continue.
Questions
Questions about the service.
Can we buy training without running the census first?
No. The census takes 18 minutes per person and costs nothing while founding places remain, however without the census we cannot say which teams need training and which teams need something else. A course for a team whose binding constraint lies elsewhere is unlikely to move that team's indicators.
What if the profile says training is not the remedy?
Then the profile will say so, and the remedy that follows will be one of the other three (peer demonstration and champions, the policy and rules restated for the role, or licences, time and a sponsor). Two of these three remedies cost less than training does. The profile is read team by team. As such, an organisation will usually need more than one remedy – and will seldom need all four in every team.
Who delivers the programme?
The programme is delivered by the founders, with accredited partners where an organisation already works with one. The champions are the organisation's own staff. The aim of our coaching is that the champions carry the practice from one business area to the next after we have gone.
How long does a programme take?
For an organisation of about 60 people, a programme takes about nine months from the first census to the second census, with the remedies delivered over the middle three to four months. The timing is set with the organisation around its leave and busy periods.
We already have a training provider. Can they deliver the ability remedy?
Yes, provided the cohorts are sorted by the profile and the indicators are fixed before the training starts. The cohorts are sorted from the profile and the indicators agreed with the provider before the sessions start, and the result is measured at four and twelve weeks in the same way as for a module we deliver.
Next step
The first census comes before any remedy.
The administrator creates the account, adds the business areas and sends the link. Where an organisation wants to talk about a programme first, we reply within one working day.
You can also email sales@aielabs.co.uk.