---
title: An AI use case is a small reorg. Lead it like one.
description: AI use-case lists score the tool. The work it changes goes unscored and turns into org design debt. How to find out what a use case asks of your team.
---

[Blog](https://managementkits.com/blog)

# [An AI use case is a small reorg. Lead it like one.](https://managementkits.com/blog/ai-use-case-org-design-debt)

 Written by [Olaf Bach](https://managementkits.com/blog/author/olaf-bach) | 18 Nov 2026

**Most AI use-case lists are scored like a purchase: what the tool costs, how many hours it saves, how hard the integration is. But a use case that is non-trivial (because it exceeds the scope of a single human job role and its tasks) typically changes who decides, how work is handed over and what the job actually is. In short, it requires a small reorganization. If you delay this work, you run up what we call org design debt, and that debt stands between you and the real value of your AI investment.**

The list usually comes as a spreadsheet. Thirty or forty use cases, collected in workshops, each with a row. The columns say how many hours a year the case would save, what the licences cost, whether the data is ready and which risk class applies. Plot it in a matrix, do some portfolio considerations, and the top five become this year's priorities. I have helped build lists like this, and they do their job: they make a crowded field discussable, and they stop the loudest sponsor from winning by default (even though I would always consider self-selection of leaders in picking the next in line).

What they score, though, is (projected capability of) the tool. Take a case near the top of many lists: an assistant that routes and drafts the initial replies to incoming customer requests. The hours are easy to estimate. To get them, somebody has to decide which drafts may go out unread, and who answers for the ones that were wrong. The second pair of eyes moves from every reply to a sample. The experienced people stop writing and start handling exceptions, which is a different job (and not everybody wants it). And somebody has to say what the saved time is for. None of this has a column.

## Why is the part nobody scores the bigger part? Or: why do some firms get so much more out of their tech than others?

Here our soft spot for classic research is useful. Twenty-six years ago, two economists asked why some firms got so much more out of their computers than others. Erik Brynjolfsson and Lorin Hitt reviewed the firm-level evidence and found that the technology was the small item ([Journal of Economic Perspectives, 2000](https://doi.org/10.1257/jep.14.4.23)). They cite a study of large enterprise-software projects in which hardware made up less than 4 percent of a start-up cost of about 20 million dollars, and software another 16 percent. The rest went into redesigning processes and training people. Their own **estimates put the organizational assets a firm builds around its computers at several times the value of the computers, possibly ten to one.** The firms that made these investments were the productive ones.

*A side note from the paper: a manufacturer the authors call MacroMed bought flexible machines to offer more variety and was disappointed, because the workers kept an old rule never to stop a machine for a changeover. The gains came when the machines arrived together with flat pay, operators who scheduled their own machines and fewer layers of management.*

**An AI use case is a small reorganization, and it should be managed like one.** The AI use case approval is the entry ticket. What the team has to change is the investment, and the return comes from that investment.

*Working on this in your own organization? [See how the MK Practice works →](https://managementkits.com/practice?utm_source=blog&utm_medium=cta&utm_campaign=p1-practice&utm_content=pn07-mid)*

## What happens to the AI use case list when you count the organizational change?

The quick wins lose some of their shine. A case that asks nothing of the team is cheap for a reason: nothing around it moves, so the minutes it saves are refilled by the next task, as we described in [the Note on saved time](https://managementkits.com/blog/time-saved-by-ai). Such cases are welcome conveniences. Their contribution to the business case is another matter.

The cases with substantive org design debt move up, provided the effort is one your team can carry this year. If a case needs a decision right that sits two levels above you, it is a question for the programme and belongs there. If it needs changes you can make inside your own team, [it is yours](https://managementkits.com/blog/where-ai-transformation-starts), and it is probably worth more than its row suggests.

The measure changes too. Hours saved is still the most common yardstick: in this year's survey of large US companies by Wharton and GBK Collective, 40 percent of departments used it, more than any other measure ([reported in Forbes, October 2026](https://www.forbes.com/sites/gabrielalinzainescu/2026/10/13/strong-ai-roi-is-nearly-3x-more-common-when-firms-track-growth/)). However, I dare to say this figure speaks to a widespread lack of imagination. Departments that tracked at least one growth measure reported strongly positive returns far more often than those that tracked efficiency alone, 35 against 13 percent. That is self-reported and proves no cause.

But it does fit the old finding: the return comes as better output for the customer (more variety, faster answers, better service) rather than as hours saved. And you can only show it if you wrote down where the team stood before the tool arrived.

## How do you find out what a use case asks of your team?

Some ideas & suggestions (admittedly high level and details will vary and have to be tailored to specific context)

- Start with the case at the top of your own list, and assess before you decide.
- Follow **one real piece of work** through the team as it runs today, and mark every point where the tool would sit.
- At each point, note what would have to be different for the saving to be real: [a decision someone else takes](https://managementkits.com/blog/who-may-overrule-the-ai) (who may send a draft unread?), a check that moves (from every reply to a weekly sample), a task that ends (nobody writes first replies any more), a skill that is missing (judging a draft is not the same as writing one).
- Put the notes next to the row. If the list is short, do the same for the second and third case and compare.

Then go back to the people who built the list and ask for one more column. It does not need a number. One sentence per case on what the team has to change, and a name for who leads that change, is enough to alter the discussion about priorities.

This is also where the MK Practice begins. Before any step is planned, we sit down with you for a working session: your view of the problem next to our diagnostic of your team. Together we record where the team stands, so that six weeks later there is a before and an after you can show. If one of the cases on your list is about to land in your team, [this is how the six weeks work](https://managementkits.com/practice?utm_source=blog&utm_medium=cta&utm_campaign=p1-practice&utm_content=pn07-close).

The spreadsheet can stay. It needs one more column, and that one has to be filled with the small reorg in mind, so to keep your org design debt at bay.

You don't read a practice. You run it.

Bringing AI into how a unit works is an organization transformation, and someone has to lead it. In the MK Practice you lead it on your own unit, in your real work, with a practitioner who reads every step and signs it. Your employer can pay for it.

[See how the MK Practice works →](https://managementkits.com/practice?utm_source=blog&utm_medium=cta&utm_campaign=p1-practice&utm_content=pn07-org-design-debt)

*Want to talk it through first? [Book 20 minutes with a Management Kits team member](https://calendly.com/managementkits-demo/talk-first?utm_source=blog&utm_medium=cta&utm_campaign=p1-practice&utm_content=pn07-call). Running it with a whole leadership team? [See the MK Practice for organizations](https://managementkits.com/enterprise-solutions?utm_source=blog&utm_medium=cta&utm_campaign=p1-practice&utm_content=pn07-org).*

[View full post](https://managementkits.com/blog/ai-use-case-org-design-debt)

```json
{
  "@context" : "http://schema.org",
  "@type" : "BlogPosting",
  "author" : {
    "@type" : "Person",
    "name" : "Olaf Bach"
  },
  "dateModified" : "2026-11-18T20:50:05.581Z",
  "datePublished" : "2026-11-18T20:09:58Z",
  "headline" : "An AI use case is a small reorg. Lead it like one.",
  "image" : {
    "@type" : "ImageObject",
    "height" : 700,
    "url" : "https://26208959.fs1.hubspotusercontent-eu1.net/hubfs/26208959/Management-Kits_Approved-Then-Org-Design-Debt_pn07_transparent.png",
    "width" : 1200
  },
  "mainEntityOfPage" : "https://managementkits.com/blog/ai-use-case-org-design-debt",
  "publisher" : {
    "@type" : "Organization",
    "logo" : {
      "@type" : "ImageObject",
      "height" : 60,
      "url" : "/hs/hsstatic/content_shared_assets/static-1.4092/img/default-amp-logo.png",
      "width" : 60
    },
    "name" : "Blog"
  }
}
```