An AI governance charter is written almost entirely in the grammar of seats. Voting members and advisory members. Escalation to the seat above. Quorum expressed as a count of chairs occupied. Meeting cadence attached to a standing invitation rather than to anybody's diary in particular. Every noun in it describes a position, and positions are exactly the thing an enterprise reorganises. Give it a reorganisation or two and the committee still exists on paper while the people who argued it into existence are doing something else.
The executive committee governance programme I worked on for a consumer and SME banking group operating across several European markets schedules its charter the way most good ones get built, in the room, in ninety minutes, down to the operating detail: mandate, membership, the split between voting and advisory roles, escalation paths, cadence and quorum. What is more interesting sits two slides later, in the follow-up plan. In the fortnight that turns the sessions into a final report and a roadmap, the plan calls for one further artefact, and it is the only one in the whole document that carries a person's handwriting. Signed commitment cards, one per executive, to be collected from all of them.
A charter binds a seat and a card binds a signatory
The programme scoped eleven decision sessions, each with a named output artefact as its exit condition, across roughly fourteen hours of room time. Session four produces a draft charter. Session seven produces a nominee list of ten to fifteen AI champions across departments, chosen on mindset and influence rather than rank, nominated cross-departmentally and then voted on. Both artefacts distribute governance. Only the second one names humans, and it names them as a class rather than as signatories.
The commitment card is the third move, and it does something neither of the other two can. A charter obliges whoever occupies the chair. A champion list obliges a group. A signed card obliges the individual who wrote their name on it, and it keeps obliging them after the committee is renamed, after the reporting line moves, after the sponsor changes. Whether that durability is why the device is in the pack, the pack does not say. It is my reading of what the artefact is for.
The old-fashionedness is the mechanism. An email acknowledgement records receipt. A name in a responsibility grid records somebody else's intention about you. A card you signed by hand is the only one of the three that required a deliberate physical act from the person being bound.
The signature sits in the documentation band
The follow-up plan runs in three dated bands. An immediate band at forty-eight hours: circulate decisions, action items and resources, launch a participant feedback survey, hold a separate detailed debrief with the sponsor. A planned band in weeks one and two: the final report, gated on completion of the full session series rather than any single session, plus a detailed action plan, a thirty-sixty-ninety roadmap, and the signed cards. A recurring band with a hard deadline at the end of week two: the thirty, sixty and ninety-day review dates locked in calendars, with their agendas already written.
Where the cards sit in that sequence is the finding. The obvious place to collect a signature is the close of the workshop, while the room is warm and everyone has just agreed with each other in public. This design does not do that. It puts the signature a week or two downstream, in the band whose other three items are all documents. Whether each executive reads the final report before signing, the plan does not state. What it states is that the signature is gathered among documents rather than at the close of a session.
That placement costs the emotional peak; a card signed a week or two later is signed in a normal week, by someone who has since remembered their other commitments. What it buys is a signature against text rather than against atmosphere. A commitment made at the close of a session is a commitment to a feeling about a session. A commitment collected in the same band as the final report and a dated roadmap is a commitment made against clauses the signatory has had a chance to dispute, and it is worth more because it is harder to get.
The same logic runs on the way in. The programme's preparation contract, which I wrote about in Before You Book the Executive AI Workshop, Budget Ten Hours of Pre-Work, treats the room as a rendering surface for work done beforehand. The card treats the signature the same way: it is the rendering of a decision whose text already exists, not the moment the decision is made.
One artefact per person, everything else per programme
Count the deliverables named in the follow-up plan. Decision summary: one. Feedback survey: one instrument. Sponsor debrief: one conversation. Final report: one. Action plan: one. Roadmap: one. Review dates: three, booked once. Commitment cards: as many as there are executives.
That asymmetry is unusual. Programme documentation almost always addresses the organisation, and an organisation is a convenient place to put an obligation nobody has to carry. The card is the single point in the design where the obligation is disaggregated back down to individuals, and it happens in the same fortnight the aggregate documents are produced, so the two forms sit side by side rather than one replacing the other.
There is a second-order effect the slide does not claim, and I flag it as inference rather than source. The verb on the slide is collect. Collection implies a set, and a set implies a gap. If one card does not come back, that gap is legible, attributable and quiet, and it tells the sponsor something no aggregate completion figure ever will. A card that never arrives is not an administrative loose end. It is the most useful piece of information the follow-up phase produces.
- 1 per executive
- Signed commitment cards, specified for collection in weeks 1-2
- 3 bands
- Follow-up structure: immediate at T+48h, planned in weeks 1-2, recurring by end of week 2
- 90 min
- Room time allocated to drafting the governance committee charter, membership to quorum
- 0%
- Execution status on the follow-up plan at the time of the snapshot
Analogue on purpose, in a programme that is analogue elsewhere too
The card is not an isolated eccentricity. The same pack carries a readiness checklist for the person running the sessions, and it is deliberately physical: printed wall canvases, sticky notes and voting dots, a tested timer. That checklist is flagged as a precondition for the high-engagement sessions to work at all, not as venue logistics.
A group of senior people asked to place a dot on a wall canvas commits in front of each other in a way a poll link does not reproduce. The programme uses physical artefacts at the two points where the design needs the act to cost something, and documents everywhere else.
A charter is addressed to whoever is sitting in the chair. A card is addressed to the person who signed it, and that person carries it into the next org chart whether or not the chair survives.
What this generalises to
The pack states its own premise in one line: the real work begins after the workshop closes, and immediate action prevents strategic drift. Most programmes agree with that sentence and then respond to it entirely with institutional instruments, a charter, a roadmap, a reporting cadence. All three are necessary. None of them is answerable to anyone.
Three things make the card version work, and none of them requires a governance programme to copy. Put the personal commitment in the same band as the written record, so the signature lands against text. Make the act physical, so it costs a moment of deliberate attention. Keep a count, so a missing one is visible.
The execution meter on this plan read zero percent when the snapshot was taken, because the sessions had not finished. That is the honest state of it. What the design gets right is that it built the personal artefact into the plan at the same level of specificity as the corporate ones, with a date, a verb and a number attached, instead of hoping that a committee charter would somehow bind the people who wrote it.
Details are as recorded in an executive committee information pack and workshop snapshot prepared for a consumer and SME banking group operating across several European markets: its eleven-session schedule, its governance charter session and its three-band post-workshop follow-up plan. The commitment cards, the deadlines and the review cadence are the programme's design requirements, not measured outcomes; the follow-up plan's execution meter reads zero percent because the sessions had not completed at the time of the snapshot. Reading the signed card as the one artefact in the plan that outlives a reorganisation, and reading an uncollected card as a signal rather than an administrative gap, are ours.
“A charter is addressed to whoever is sitting in the chair. A card is addressed to the person who signed it, and that person carries it into the next org chart whether or not the chair survives.”
Get in touch
Put RealAI’s applied-AI team on your hardest data problem.
We help enterprises move from pilots to production: sovereign models, governed data, and agents you can audit. Start with a value-first assessment.