A composite, not a real client. The people and figures below are blended from patterns I meet repeatedly in my work.
Fourteen proposals, budget for three
Rami is the CFO of a 600-person distribution group. On a Thursday afternoon he has fourteen AI proposals spread across the boardroom table. Customer service wants a chatbot. Sales wants lead scoring. Finance wants invoice matching. Procurement wants supplier risk alerts. Operations wants demand forecasting. Nine more arrive from the remaining departments.
Each proposal has a slide, a champion and a reasonable argument. The budget, after a year of currency losses and one delayed receivable, covers three.
The usual move is to rank the fourteen, fund the top three and send the other eleven away with a vague "maybe next year." In my work I have watched that move produce the same result every time: eleven department heads who do not know why they lost, and who return in January with the same slide. That is the cost of an unexplained no.
Why a list of yeses is a wish list
A roadmap that lists only what the company will do tells the board nothing about the judgment behind it. Anyone can list three projects. The board cannot tell whether those three beat the other eleven or simply reached the table first.
Worse, an unexplained no is not a decision. It is a postponement with no terms, and it returns at every budget meeting, consuming the one resource the company cannot buy more of: executive attention.
The fix is a Stop List. Every excluded idea gets a reason, a category and an unlock condition. The unlock condition is an event, never a date. A date says "ask again in the third quarter," which invites lobbying as the quarter approaches. An event says what must be true before the idea returns, which turns lobbying into work.
I have written before about choosing the first use case, setting risk limits and giving approved cases an expiry date. This is the other half: publishing what was left out. The earlier pieces are on choosing a first use case, a board risk appetite statement and giving a business case a half-life.
The four stops
Every excluded idea lands in exactly one of four stops. If you cannot decide which, the idea is not yet understood well enough to be funded.
- Not yet. A dependency is missing: clean data, a documented process or a workflow owner. Test: if funding arrived tomorrow, could the team start on Monday? If not, it is a Not yet. Unlocks when: the named dependency exists and someone can show it.
- Not here. The idea is valid, but the organisation cannot absorb another change this year. Test: could the people who must change how they work do so without dropping something already promised? Unlocks when: a current programme finishes or capacity is freed in the affected team.
- Not ours to build. The capability is a commodity, and buying or waiting beats building. Test: would a competitor gain an edge from our version over a purchased one? Unlocks when: we take a buy decision, or a vendor meets a requirement we have written down.
- Not worth it. The value is below the management attention it would consume. Test: if it worked perfectly, would anyone on the executive team change a decision? Unlocks when: nothing, unless the facts change. Then it re-enters as a new proposal.
Two rules keep the list honest. First, an unlock condition must be something an outsider could verify. "When we are ready" is not one. Second, Not worth it must stay small. If a third of your list lands there, you are using it to avoid the harder sorting.
Sorting the fourteen
Back to Rami. Over two sessions with his team, the fourteen sorted like this.
- Go (3): invoice matching, collections prioritisation and routing of customer enquiries. Each has a business owner, usable data and a team ready for the change.
- Not yet (5): demand forecasting, lead scoring, supplier risk alerts, contract review and churn prediction. In each case the data sits in several systems, spreadsheets and branch-level habits that do not agree on what a customer or a product code is. The dependency is a single, trusted record.
- Not here (3): dynamic pricing, warehouse re-slotting and a CV screener. Price lists change weekly with the currency, and the sales team is already absorbing that volatility by hand. A model on top would be a change the organisation cannot carry this year.
- Not ours (2): meeting summaries and document search. Both are commodity capabilities. The unlock is a buy decision, not a project.
- Not worth it (1): a narrated dashboard for a report the board reads twice a year. It would work, and nobody would act differently because of it.
In Lebanon and much of MENA, I see the two middle stops carry more weight than they would in a calmer economy. Fragmented data is the normal condition, not a failure, so Not yet is often the honest answer. Currency volatility and constant firefighting reduce the change capacity of every team, so Not here is too. Naming this plainly is kinder than funding projects that stall.
Putting the stop list in front of the board on one page
The Stop List is only useful if the board sees it. I ask for one page with three columns.
- Do: the three funded bets, each with a named business owner.
- Stop, with unlock: the eleven excluded ideas, each tagged with its stop and the event that would bring it back.
- Owner of the next look: one named person per stop who watches for the unlock event and brings the idea back only when it happens.
A note on that owner: it should not be the idea's sponsor. A sponsor watches for the event that favours their proposal. A neutral owner watches for the event itself.
Rami's version fit on one sheet. The chairman read the eleven stops first, which surprised him less than it surprised the department heads, who could now see that their ideas had been judged and not ignored. Two of them asked what it would take to move from Not yet to Go. That is the conversation you want.
A roadmap is credible when you can see what it cost to choose
Any company can announce three AI initiatives. What earns trust is showing what those three displaced, and why. A Stop List makes the cost of choosing visible, and it gives every sponsor a fair route back that does not run through the CFO's calendar.
If you hold more proposals than budget, do one thing this week: take your longest list of requests and write one stop and one unlock event beside every item you will not fund. Then ask which stop holds your most requested idea.
Related evidence: Nielsen Norman Group's chatbot research found that users generally reacted favourably when a bot owned its failure and offered an escape hatch, such as a phone number or a live agent. (Nielsen Norman Group research on chatbot UX)
Article 14 of the EU AI Act requires high-risk AI systems to be designed and developed, including with appropriate human-machine interface tools, so that natural persons can effectively oversee them throughout the period they are in use. (EU AI Act Article 14 on human oversight)
