# Chapter 12: The first 90 days

> Every chapter in this playbook feeds one plan. In 90 days, a leadership team can choose its use cases, prove the first one on real data and set up the foundations to scale.

The AI-Native Playbook for Enterprises · Twistag · 6 min read · Updated: 2026-10-09

## In one minute

1. **Decide in month one.** A short list with owners, baselines and money figures, and one use case chosen to prove first.
2. **Prove in month two.** A working agent on real data, inside the real process, with a person approving what matters.
3. **Set up to scale in month three.** The decision on the first agent, the business case for the next, and the register, tiers and roles that every later agent will use.

## Why 90 days

Ninety days is long enough to prove real value and short enough to keep attention. MIT's 2025 research found that mid-market companies that succeeded with AI moved from pilot to full roll-out in about 90 days, while large enterprises took nine months or more and converted fewer pilots. Speed was one of the clearest differences between the companies that got value and the rest.

The plan below brings together the earlier chapters. It assumes a leadership team that has decided to act, a sponsor with the authority to move budget, and a small amount of outside or internal help to build.

## Month one: decide

**Weeks 1–2. Choose the work.**

- Each executive names the three processes that cost their area the most time or money (Chapter 1).
- A two-hour leadership session scores every candidate on value, effort and risk, with finance in the room, and agrees a short list of three to five.
- Each item on the short list gets an owner and a baseline figure.
- Leadership picks one to prove first, using the four tests in Chapter 3.

**Weeks 3–4. Prepare the proof.**

- Map the chosen process with the people who run it, and label each step as rules, judgement or relationship (Chapter 2).
- List every data source the agent needs, request read access, and run the agent readiness check (Chapter 4).
- Decide buy, build or partner for the first agent (Chapter 5).
- Write down the target, the stop criteria and the date of the decision (Chapter 3).

By the end of month one, the company has a short list, a baseline for each item, and a first proof ready to start on real data.

## Month two: prove

**Weeks 5–8. Build on real data.**

- Build the agent against real documents and systems, tested every week against a set of real past cases.
- The people who run the process review it every week.
- Tracing, cost tracking and a way to pause the agent are built in from the start (Chapter 7).

**Weeks 9–10. Run with approval.**

- Put the agent into the live process in assist or approve-by-exception mode (Chapter 2).
- Measure the same numbers as the baseline, on live volume.

In parallel, the central team starts the agent register and places every AI system already in use in a governance tier (Chapter 8).

## Month three: decide and set up to scale

**Weeks 11–12. Measure and decide.**

- Bring value, quality, cost to run and risk to the owner and the leadership team (Chapter 10).
- Make the call: scale, change or stop.
- If scaling, write the one-page business case with low, expected and high cases, agreed with finance (Chapter 6).

**Week 13. Plan the next stage.**

- Choose the next two use cases from the short list.
- Name the owner, operator and engineer for each live agent, and add the roles to job descriptions (Chapter 9).
- Agree the quarterly review rhythm with the board.
- Decide whether shared services, such as model access and cost tracking, are needed yet (Chapter 11).

**Free worksheet: 90-day plan** (PDF and spreadsheet). A one-page plan with owners, budget ranges and milestones for each of the 13 weeks, ready to fill in during a leadership session. Request it by email: https://twistag.com/guides-and-playbooks/ai-native-playbook/chapter-12#worksheet

## What day 90 looks like

A good first 90 days ends with:

- one agent working in a real process, with measured results against a baseline;
- a decision, taken with numbers, on whether to scale it;
- a short list of the next use cases, each with an owner and a baseline;
- a business case for the next stage that finance has agreed;
- a register of every AI system in use, each in a governance tier;
- named owners and operators, and a quarterly review in the calendar.

That is enough to fund the next stage with confidence, and to show the company that AI is changing how work gets done.

## A note on feasibility first

Some use cases carry more uncertainty: a new data model, many systems, or a product that does not exist yet. For these, a short feasibility phase before the build reduces the risk.

Source.app, a marketing attribution platform Twistag built with the client's team, started with a six-week technical feasibility phase covering the data model, the integration strategy and the event schema, before any engineering commitment. The platform went on to integrate six marketing tools into one normalised model, and went from concept to a live first version, used by early enterprise customers, in 11 months, just under its target.

Where we saw it: [A production marketing attribution platform in 11 months](https://twistag.com/case-studies/source-app-marketing-attribution-platform) (Source.app)

For most first agents, the proof in month two does the same job in less time.

## Who does what

| Who | Month one | Month two | Month three |
|---|---|---|---|
| CEO or sponsor | Sets the goal, chairs the scoring session | Removes blockers, protects the team's time | Chairs the decision, approves the next stage |
| Finance | Agrees baselines and value logic | Tracks build costs | Signs the business case |
| Business owner | Owns the first use case and its baseline | Joins every weekly review | Presents the results and the decision |
| Central team | Prepares the session and the short list | Starts the register and tiers | Sets up roles and the review rhythm |
| Build team | Maps the process, checks the data | Builds and runs the agent | Hands over to the running team |

## Getting help

Many companies run the first month with outside help, so the short list and the first proof are shaped by people who have done it before. Twistag's AI Readiness Sprint covers month one in two to four weeks, and its teams build and run the proofs that follow. Whatever the route, the company should come out of the 90 days owning its short list, its data, its agents and the skills to run them.

## Common mistakes

- **Starting with a strategy document.** Months spent on a vision paper, with no agent at the end.
- **Too many proofs at once.** Five teams, five pilots, and none with the attention to succeed.
- **No decision date.** The first proof drifts past day 90 with no call made.
- **Skipping the foundations.** The first agent succeeds, but there is no register, no roles and no rhythm, so the second one starts from scratch.
- **Keeping it quiet.** The rest of the company never hears about the results, and the next team starts with doubts.

## What to do this month

1. Name the sponsor and the start date for the 90 days.
2. Book the two-hour scoring session with the leadership team and finance.
3. Ask each executive to bring their three most costly processes, with rough figures.
4. Request the 90-day plan worksheet and fill in the owners before the session.

## Questions for the leadership team

1. Who in the leadership team owns the 90-day plan, and when does it start?
2. Which executive will own the first agent, and has their team set aside time for it?
3. What decision will leadership make on day 90, and what numbers will it need?
4. What budget is approved for the first 90 days, and what is held back for the next stage?
5. How will the rest of the company hear about the plan and its results?

## Sources

1. [The GenAI Divide: State of AI in Business 2025](https://virtualizationreview.com/Articles/2025/08/19/MIT-Report-Finds-Most-AI-Business-Investments-Fail-Reveals-GenAI-Divide.aspx), MIT NANDA, 2025-07
2. [Applied AI Index 2026](https://www.bcg.com/press/30september2026-ai-starting-to-pay-off-companies-generate-value), Boston Consulting Group, 2026-09

---

Source: https://twistag.com/guides-and-playbooks/ai-native-playbook/chapter-12