High-Impact Use Cases

The reality

A founder runs a 28 person property management business and has, across the last year, signed up for six AI tools the team has tried at various points: a meeting transcription service, a chat assistant for tenant queries, a maintenance scheduling AI, a marketing copy generator, a financial reconciliation tool, and a contract reviewer. Total spend across the year was AED 22,000 (USD 5,990). Total hours saved, measured honestly, was roughly 40 hours, mostly from the meeting transcription. Every other tool had been tried for two to four weeks and quietly dropped. The cost extended well beyond the AED 22,000 (USD 5,990). It was the team's growing sense that AI was overhyped, the founder's growing fatigue with the category, and the missed opportunity to deploy two or three tools well instead of six tools poorly. AI use cases in a service business come from identifying the three to four workflows where AI genuinely fits, deploying carefully, measuring honestly, and ignoring the next demo.

Read this if

• The business has tried more than three AI tools in the last year and cannot point to clear hours saved
• AI use cases are picked based on what was demoed at the last conference, with no link to what the business actually needs
• A team member has built a personal AI workflow that saves them time, but the workflow has not been shared or institutionalised
• The founder cannot name the top three workflows in the business where AI would produce the largest return
• AI tool spend has grown to more than AED 1,500 (USD 410) per month with no defined return measurement
• The team treats AI experimentation as a side project, with no acknowledgement that it is a system change

What dysfunction costs

Tool sprawl cost. Six AI tools at AED 200 to AED 600 (USD 54 to USD 163) per month per tool, used by two or three people each, costs the business roughly AED 15,000 to AED 30,000 (USD 4,085 to USD 8,165) per year. The cost is the spend plus the cognitive load of the team trying to remember which tool does what.

Missed opportunity cost. A business that experiments randomly across six tools rarely deploys any of them well. The opportunity cost is the two or three tools that, if deployed properly, would have saved 200 to 400 hours of senior time across a year.

What success looks like

When AI use cases are designed:

• The business has identified three to four workflows where AI provides clear leverage and is deploying against those, ignoring the next demo
• Each AI use case has a named owner, a measured baseline (hours per week before AI), and a measured outcome (hours per week after)
• New AI experiments go through a defined gate: does this fit a top three to four workflow, or is it a side experiment with a defined budget and timebox
• Personal AI workflows that save individual team members time are surfaced quarterly and considered for institutionalisation
• AI tool spend is reviewed quarterly against measured return
• The team understands that AI deployment changes how the work happens, well beyond the subscription decision

The framework

AI use case selection runs as four layers. Each layer answers a different question about where AI fits in a service business.

Layer 1: The four high-leverage categories

Most service business AI use cases that produce real return fall into one of four categories.

Document and content generation at scale. Drafting client proposals, writing first-draft contracts, generating onboarding documents, producing first-draft project reports. The leverage is real because these tasks consume senior time and AI produces an 80 percent draft a senior person can edit in 20 percent of the time it would have taken to write.

Search and retrieval across the business's own files. RAG systems (covered in RAG: The AI That Reads Your Own Files) that let the team query the business's own documents (contracts, project history, client records) in natural language. The leverage is real for businesses with significant document volume.

Voice and meeting capture. Transcription, summarisation, and action capture across meetings and client calls. The leverage is real because senior team meetings consume 5 to 15 hours per week and the cognitive load of capture is significant.

Structured analysis of unstructured data. Reading client emails for sentiment patterns, categorising support tickets at volume, extracting structured data from invoices or contracts. The leverage is real for businesses with significant volume on the input side.

The behaviour to adopt this week: which of the four categories has the largest fit for this business? Pick one or two. The other two get parked.

Layer 2: The fit test

For each potential use case, run the fit test. Five questions. (1) Is this a workflow that consumes 5 or more senior hours per week? (2) Is the input mostly text, voice, or structured data that AI handles well? (3) Is the cost of an AI error low to moderate, or can a human review catch errors before they reach the client? (4) Is there a measurable baseline (hours per week, accuracy rate) we can compare against? (5) Is there a named owner who will run the deployment for at least 90 days?

A use case that fails on any of the five questions is not the right starting use case. A use case that passes all five is a candidate for deployment.

The behaviour to adopt this week: pick the top three potential use cases. Run the fit test on each. Notice which one or two pass cleanly.

Layer 3: Deploy with measurement

The deployed use case has a named owner, a 90 day pilot, and a defined measurement. Hours saved per week. Output quality reviewed. Team adoption rate. The pilot ends with a structured review: did the use case deliver, what is still broken, do we expand or kill.

When deployment runs without measurement: the use case becomes a felt-impact, with no measured-impact behind it. The founder's instinct that "AI is helping" cannot be tested. The next tool that promises the same thing gets adopted on the same instinct.

The behaviour to adopt this week: for the chosen use case, write the 90 day plan. Owner, baseline, weekly review, end-of-90-day decision criteria.

Layer 4: The quarterly review and the gate

Once a quarter, the business reviews AI use cases against measured return. Tools that delivered get expanded. Tools that did not get killed. New experiments are accepted only if they fit one of the top three to four workflows or have a defined small budget and timebox as a side experiment.

The gate prevents the next demo from becoming the next abandoned subscription. AI tool spend stays within a defined annual budget. The team's attention stays focused on the workflows where AI is producing real return.

The behaviour to adopt this week: schedule the first quarterly AI review. 60 minutes. Standing template: tool, use case, measured return, decision (expand, hold, kill).

A founder you might recognise

A founder runs a 26 person legal services firm in DIFC. AED 13M (USD 3.5M) last year. Across 2024 and 2025 the team had tried 11 AI tools across various workflows. By Q4 2025 the team had broadly given up on AI as overhyped. AI tool spend was around AED 18,000 (USD 4,900) annually with unclear return.

In Q1 2026 the founder ran the four layer reset. The four high-leverage categories were reviewed against the firm's actual workflows. Document generation (drafting first-draft contracts, briefs, and client memos) was the clear top use case. Search and retrieval across the firm's own document library was the second. Meeting capture for client calls was the third.

The fit test was applied to each. All three passed. The team killed seven of the 11 existing tools, kept one (a meeting transcription service that was working), and added two: a document drafting workflow with a paid LLM and a small RAG deployment over the firm's contract library. Each had a named owner, a 90 day pilot, and a measurement plan.

By the end of Q3 2026 the document drafting workflow was saving roughly 12 senior hours per week. The RAG deployment was saving 4 hours per week and reducing one specific source of error in contract review. The meeting capture continued to save 3 hours per week. Total AI tool spend had dropped to AED 11,000 (USD 3,000) annually. Total measured hours saved was 19 per week, or roughly 950 per year. Each senior hour was costed at roughly AED 250 (USD 68), which valued the saving at AED 237,500 (USD 64,640) annually. The cost of the reset had been four hours of senior team review time. The output had been a business now extracting real return from AI in three workflows, replacing diffuse experimentation across 11.

Working through it

1. Review the four high-leverage categories against the business. Document generation, search and retrieval, voice and meeting capture, structured analysis of unstructured data. Which one or two have the largest fit?

2. List potential use cases against the categories. For each, run the fit test (5 questions). Pick one or two that pass cleanly.

3. Audit the existing AI tool stack. Which tools are delivering measured return? Which are not? Kill the ones that are not.

4. Deploy the chosen use case with measurement. Named owner, 90 day pilot, baseline, weekly review, end-of-90-day decision criteria.

5. Install the quarterly review gate. New AI experiments must fit a top use case or have a defined small budget and timebox. The gate keeps the team focused.

Common mistakes

• Picking use cases based on demos instead of fit. A demo that looked great at a conference rarely produces the largest return for the business. Run the fit test.
• Skipping baseline measurement. A use case without a baseline cannot be evaluated. The hours saved per week or the accuracy rate is the comparison the 90 day review hinges on.
• Letting individual team members run personal AI workflows without surfacing them. A team member who has built a personal AI workflow that saves them 5 hours a week is producing leverage that is not yet institutional. The quarterly review surfaces these.
• Deploying more than two AI use cases at once. A team running two AI deployments in parallel can hold both. Three or four splits attention. Sequence the deployments.
• Treating AI deployment as a tool subscription. A subscription pays for access. A deployment changes a workflow. The deployment is what produces the return, not the subscription.

Self-assessment

Y or N for each.

1. Has the business identified the top three to four workflows where AI provides clear leverage, against the four high-leverage categories?
2. Does each AI use case have a named owner, a measured baseline, and a measured outcome?
3. Are new AI experiments gated against the top use cases or a defined small budget and timebox?
4. Are personal AI workflows surfaced quarterly and considered for institutionalisation?
5. Is AI tool spend reviewed quarterly against measured return?
6. Has the team killed at least one AI tool in the last year that did not deliver measured return?
7. Can the founder name the measured hours saved per week from the top AI use case in the business?

Five or more "yes" answers means AI use case selection is doing the work it is supposed to do. Three or four is the band where the discipline exists in part but tool sprawl is still consuming attention. Two or fewer means the next AI tool will follow the same pattern as the last six.