# Shareholder Pressure on AI: The Questions Management Must Answer

Author: Marin Niehues, Niehues Consulting · Published: 2023-11-23 · AI Strategy & Business Case
Source: https://www.niehuesconsulting.de/en/blog/shareholder-pressure-ai/

When shareholders ask about your AI strategy, you answer eight questions on objective, benefit, cost, ownership, data, risk, timing as well as stop criteria, each backed by evidence from your own business. Shareholders typically want assurance that the company is seizing the opportunity while protecting its capital and reputation. A management team that settles these questions before the meeting turns pressure into a mandate with a sequence.

The occasion is easy to name. ChatGPT has been publicly available for almost a year, the large providers announce new models and features almost weekly, and even the leadership crisis at OpenAI over the past few days has shown how quickly the environment moves. Anyone who runs a company with private investors or a family shareholder base that reads the news will be asked sooner or later: what are we doing about AI? This article shows how to prepare for that conversation and which answers hold up.

**Shareholders ask about AI when they see competitors acting and miss clarity about their own company's course. The best answer is a short plan with a target figure, a named owner as well as a decision date.**

## What do shareholders expect from management on AI?

Shareholders expect management to take a reasoned position that can be measured against the company's results. Three interests sit behind the question. The first is value creation: where does AI improve revenue, margin or lead time by an amount that shows up in the accounts? The second is protection of the existing business, meaning the concern that a competitor with a better cost structure or faster service will win market share. The third is control of risk, because nobody wants to learn from the press that employees have pasted confidential documents into a public tool.

These interests match what every investment decision demands, which helps. In this respect AI is a project like any other, with capital outlay, uncertainty as well as a point at which someone decides whether to continue. Management gains credibility by presenting the topic in that language. Arguing with model names and parameters tells shareholders that the topic has not yet been translated into the numbers.

The pressure itself is legitimate. Employees already use the tools, customers ask questions and competitors publish first announcements. A company that develops no position in the first year after large language models arrived loses learning time. How to tell announcements from real progress is covered in the article on the [AI hype cycle](/en/blog/ai-hype-cycle-progress-versus-announcement/), and how the fear of missing out arises is explained in the article on [AI FOMO in companies](/en/blog/ai-fomo-companies-investing-wrongly/).

## Which eight questions must management answer?

Management must answer eight questions, and each can be supported with evidence from the company itself. The table lists them in the order in which shareholders usually ask, together with the proof that makes the answer credible.

**The eight shareholder questions and their evidence**
Question | Answer in one sentence | Evidence for the answer | With Niehues Consulting
What are we trying to achieve with AI? | One metric that should change measurably within twelve months | Baseline and target value in the plan | Business outcome fixed before technology
Where is the greatest benefit? | Two or three applications that affect revenue, cost or lead time | Process mapping with time and volume per transaction | Sequence by value and feasibility
What will it cost? | Total cost of introduction, operation as well as changed workflows | Twelve-month cost model with stated assumptions | Business case confirmed with real values
Who is accountable? | One named owner per application in the business unit | Role description and decision path | Roles, decision paths as well as interfaces
Which data do we need? | The data sources, their quality and access | Sample test on real data | Prototype on real data in the real workflow
Which risks do we accept? | Consequences of errors, confidential data, dependence on one provider | Risk rating per application with a safeguard | Threshold before ranking
When will we see results? | A first measurable effect 90 days into live operation | Timeline with milestones and a measurement point | Metric at 90 days
When do we stop? | Stop criteria fixed before the start | Written thresholds | Decision date for continuing or stopping

Two of the eight questions deserve particular attention because they most often go unanswered. The first is the question of total cost. The licence for a language model or an assistant is usually the smallest item. Added to it are integration with systems, data maintenance, user training as well as above all the working time of the business units that redesign the workflow. The second is the question of exit. A project without stop criteria agreed in advance rarely ends, because every interim finding is read as a reason for another round.

## How do you calculate the benefit for the shareholder conversation?

You calculate the benefit from volume, time per transaction as well as cost rate, for one clearly defined workflow. The following example is illustrative and shows the calculation. Suppose a sales team prepares 400 quotations a year, each takes three hours on average and the internal hourly rate is 70 euros. The annual cost of quotation preparation is then 84,000 euros. If an assistant cuts handling time by a third, the potential is 28,000 euros. What counts is what the business unit actually redeploys, for instance on additional customer conversations. A realistic approach books only half and tests the other half in operation.

This route has the advantage that every assumption can be checked. Shareholders can compare the number of quotations, the handling time and the hourly rate with their own experience, and management can confirm them with the business unit. The result is a business case that can be defended in the room. Avoid blanket estimates of the form "ten per cent efficiency gain across the company", because they have no owner and no measurement point.

**Calculation in five steps**
- Choose a workflow that occurs often and can be counted
- Collect annual volume, time per transaction as well as cost rate
- Estimate the time saved per transaction with the business unit, then halve it
- Set one-off costs against running costs over twelve months
- Fix the measurement point and the owner before the project starts

## How should management prepare for the first conversation?

Management prepares by writing a single page with three parts before the meeting: the situation, the proposal and the decision date. The situation describes where AI is already used in the company, officially or unofficially, and which two or three applications promise the greatest leverage. The proposal names one or two initiatives with owner, cost as well as target value. The decision date fixes when the shareholders decide on continuing or extending, usually after a first prototype in the real workflow.

A survey of current use also helps. Many management teams discover that teams have already opened their own accounts with providers. That finding undercuts the argument that the company can still wait, and it supplies a first risk item for the agenda. A simple policy on handling confidential data is then a sensible first resolution that shareholders will support quickly. Showing that dependence on a single provider has been considered takes the edge off the discussion, because switching between models should be foreseen in the architecture.

The cost of external support is a frequent follow-up. The scope and terms of an engagement are discussed in the free initial consultation. For the internal discussion, the cost frame of your own staff time, which the calculation above provides, is sufficient at first.

## Which mistakes do management teams make under shareholder pressure?

Under pressure, management teams make three mistakes that can be avoided. The first is activism: ten pilots start at once, no business unit has capacity for the roll-out, and after six months there is a collection of demonstrations but no live service. The second is the shift into technology, where the discussion moves to providers, models as well as platforms while the question of business outcome stays open. The third is the commitment without a calculation, for instance a target for efficiency gains that nobody has linked to a workflow.

The same discipline counters all three: few initiatives, each with an owner, a figure as well as a date. In this way management stays able to act without denying the pressure. The article on [AI activism](/en/blog/ai-activism-board-pressure-into-plan/) shows how pressure can be turned into a sequence, and the article on [ChatGPT in the enterprise](/en/blog/chatgpt-in-business-sober-review/) describes what a year of experience with the tool has shown in operations.

## Clarifying shareholder questions on AI with Niehues Consulting

[Niehues Consulting](/en/) is a consultancy for AI management and organisation design from a delivery perspective, based in Düsseldorf. Founder Marin Niehues brings AI into operations, on projects ranging from VC-funded scale-ups to DAX 40 groups and in more than 20 countries on six continents. The answers to the eight questions arise in the AI management service line: business analysis, requirements, the business case as well as the selection and sequencing of applications.

The approach follows the Niehues Delivery Model with three phases. Three of its test questions matter most for a conversation with shareholders: which business outcome should change measurably? Is the business case confirmed with real values? Which metric shows after 90 days that the solution is landing? Anyone who can answer these three questions enters the meeting with a plan that shareholders can examine. The entry point is a free initial consultation of 30 minutes, which you can arrange via the [contact form](/en/#kontakt).

**Conclusion**
Pressure from shareholders is an opportunity to organise AI as an investment with a target figure, an owner and a decision date. Eight questions are enough for a sound conversation, and the calculation for a single workflow supplies the number that is missing in the room. A management team that presents one or two initiatives with cost, owner as well as stop criterion before the meeting keeps control of the topic.

## Frequently asked questions

### What do shareholders want to know about AI from management?

Shareholders want to know which business outcome AI improves, what it costs and who is accountable. They also ask about risk and about when the benefit will show. The eight questions in this article cover the usual expectations.

### How do I explain an AI strategy to shareholders without jargon?

You explain the AI strategy through results: one metric, one workflow, one owner as well as one date. Model names and providers belong in an appendix. An example from your own business, such as quotation preparation with volume, time and cost, convinces more than any market overview.

### How much budget should a company plan for starting with AI?

A sensible budget follows the one or two initiatives of the start, not a percentage of revenue. Calculate the cost of introduction, operation as well as business-unit working time over twelve months. The scope and terms of external support are discussed in the free initial consultation.

### Does a mid-sized company need to invest in AI now?

A mid-sized company should now develop a position and gain first experience, but it does not need a large investment. One or two initiatives with a measurable effect after 90 days are enough to learn and to give shareholders a sound basis for decision.

### What should a company do if employees already use AI tools privately?

Establish which tools are used with which data, and agree a simple policy on handling confidential information. Then offer an approved access route so that use becomes visible and manageable. Legal questions in individual cases are for specialist lawyers; this article does not constitute legal advice.

### How long does it take for AI to pay off in a company?

A first measurable effect in a clearly defined workflow is realistic within about 90 days of going live. Selection, prototype and preparation of the business units come before that. An effect on the company's overall result appears only once several applications are in regular operation.

### Who on the management team should be responsible for AI?

One member of the management team should hold overall responsibility, decide the sequence of initiatives and release resources. Each individual application also needs an owner in the business unit who is accountable for the metric. Without this two-level structure, initiatives get stuck in IT or lose their sponsor.