AI Activism: How Boards Turn Pressure into a Robust Plan

Boards turn AI pressure into a robust plan by choosing a few tasks, fixing a measurable metric and an owner for each, and deciding after 90 days on the basis of measured values. Activism shows when activity takes the place of decision: many pilots, a working group, a tool purchase without an owner. This article describes the path from pressure to plan with a decision paper, a worked example and test questions.

The pressure is real. Since the end of 2022, shareholders, customers, employees and competitors have been asking what the company intends to do with AI, and every week brings announcements that make an answer look more urgent. A board that wants to show activity in this situation has good reasons: doing nothing risks losing ground, and doing the wrong thing too early burns budget. The way out is a plan that is fast enough to answer the pressure and precise enough to avoid misinvestment.

How does a board recognise AI activism?

A board recognises AI activism by five features that often occur together in practice. The first is a multitude of parallel pilots without shared selection criteria, because every area starts its own experiments. The second is a tool that has been bought or approved before anyone named the task it is meant to solve. The third is a working group or task force without budget or decision rights. The fourth is metrics that measure activity, such as the number of training sessions or user accounts, and no business outcome. The fifth is the absence of a point at which an application is continued or ended.

Activism is the understandable answer of an organisation to pressure as long as no procedure exists that turns pressure into a decision. The evidence shows why selection matters. A study presented in September 2023 by Dell'Acqua and colleagues with 758 management consultants found clear gains in speed and quality on tasks within GPT-4's capability frontier, but on one task outside it participants were 19 percentage points less likely to be correct than the control group. The same technology therefore helps with one task and harms with the next. A plan must select tasks and cannot introduce the technology as a whole. A way to separate proven effect from announcement is offered in the article The AI hype cycle: how executives tell progress from announcement.

What does a 90-day decision paper look like?

A 90-day decision paper describes on two pages which tasks the company tests, who is accountable for them and by what the board measures success after three months. It is put to the board once and called up again after 90 days with results.

The paper contains six statements. It opens with the business outcome that is to change, for example the handling time per case in customer service, and the baseline recorded before the start. The chosen task with reasons and a named owner in the business unit who decides on workflows follow. The paper closes with the budget for the first 90 days and a stop criterion fixed in advance, so that an unsuccessful attempt ends without anyone having to renegotiate it to save face.

How does the board choose the right task?

The board chooses the task by three criteria: frequency, time spent and how easily the result can be checked. Suitable tasks occur often, tie up noticeable time and produce a result that a person can judge quickly, such as drafts of standard letters, summaries of documents or suggested replies in customer service. Unsuitable at the start are tasks where an error is expensive and nobody can check the result promptly.

A calculation shows how the choice becomes a decision variable. The example is illustrative and uses assumptions that the business unit replaces with its own values in a real case. In customer service, 25 employees each spend ten hours a week on drafting replies. That is 250 hours a week. The field study by Brynjolfsson and colleagues from 2023 with more than 5,000 service agents found on average 14 per cent more issues resolved per hour, measured across the whole work of the agents. For the example the planning assumption is deliberately lower: the business unit expects a time saving of 10 per cent on the hours spent drafting replies, which is 25 hours a week. Over 46 working weeks and at an internal hourly rate of 60 euros, this gives 69,000 euros of time value a year. Set against it are running costs of an assumed 15,000 euros and a one-off introduction effort of an assumed 30,000 euros. In the first year 24,000 euros remain on paper. Break-even lies at about 16 hours a week, which is roughly 6.5 per cent of the time spent drafting replies.

The calculation shows two things. First, the task pays off even on a cautious assumption when effort and benefit fit together. Second, the time saved is at first only an arithmetical value. It becomes capacity only when the business unit plans the freed time, for example for more cases or better advice. The owner in the business unit therefore confirms the assumption, and the pilot checks it against the baseline.

Which questions does the board ask after 90 days?

After 90 days the board asks five questions that separate progress from activity. They belong in the paper so that the date is fixed in advance.

1. Has the agreed metric moved against the baseline, and by how much?
2. Do those involved use the application in their daily work, and how many of them regularly?
3. Were there errors that a person had to correct, and how often did they occur?
4. Who is accountable for operation after the test, and has that been settled?
5. Do the measured values justify the next budget step, or does the stop criterion apply?

The answers lead to one of three decisions: extend, adjust or end. This pattern turns pressure into a rhythm. The board can give shareholders and employees a clear account: which tasks are running, what was measured and when the next decision falls. Why investments made out of fear of missing out often go in the wrong direction is covered in the article AI FOMO: why companies invest wrongly out of fear of missing out. A look back at the first year after the launch of ChatGPT is offered by ChatGPT in the enterprise: a sober look one year after the hype.

Overcoming activism with Niehues Consulting

Niehues Consulting is a consultancy for AI management and organisation design from a delivery perspective, based in Düsseldorf. Founder Marin Niehues brings AI into operation, in projects ranging from VC-funded scale-ups to DAX 40 corporations and in more than 20 countries on six continents. In the field of AI management, the work includes business analysis, the business case, and the selection and sequencing of applications, which are exactly the steps that turn many ideas into a plan.

The approach follows the Niehues Delivery Model. Three test questions matter on the path from pressure to plan: which business outcome should change measurably? What is the baseline, and how do we measure success? Which metric shows after 90 days that the solution has taken hold? The scope and terms of an engagement are discussed in the free initial consultation, which you can arrange through the contact page.