Blog · 5 min read ·
AI Integration Plan for Small Business: A 90-Day Roadmap
Direct answer: Start with one process, not a tool. A good AI integration plan lists where AI could help, what the risks are and what to do in the first 90 days, so you move in small, checked steps.
What a plan for a small business looks like
A plan is not a shopping list of software. It is a sequence. The first job is to find one process where AI saves time or reduces errors, not to connect everything at once.
Our AI integration plan works like this. We talk to you about how work really runs, list the opportunities, rank them, check the data risks and write a 90-day roadmap. It is a fixed-price product, separate from building anything.
A plan has two parts. The first is the opportunity list. The second is the order of work: days 1 to 30, 31 to 60 and 61 to 90. It does not assume every process needs AI. Some should stay manual because the risk is high or the effort is too big.
Four questions before you start
- Which process repeats often? Look for work with a pattern: email replies, invoice checks, meeting notes, product descriptions, booking confirmations.
- What happens if the AI is wrong? It can be. If the result goes to a customer, a person must approve it first.
- What data is involved? Customer records, personal data and financial data need extra care.
- Who checks the output? Name one person who reviews it before it leaves the business, and give them the time to do it.
Impact versus effort
Score each opportunity on two things. Impact is time saved, errors removed or faster decisions. Effort is setup time, data clean-up and training.
| Opportunity (hypothetical) | Impact | Effort | When |
|---|---|---|---|
| Drafting routine customer email replies | High | Low | Start first |
| Summarising notes from sales calls | Medium | Low | Start early |
| Rewriting product descriptions in bulk | Medium | Medium | Test on a small batch |
| An agent that edits order records on its own | High | High | Later |
This table is an example. A real list comes from your own processes. Start with the low-effort, high-impact item and leave complex agent work for later.
Data and GDPR risks in plain words
Many AI tools send text or customer records to an outside provider. That transfer is a risk you have to look at. The GDPR applies whenever you process personal data in the EU, and the plan looks at it for each opportunity. Ask:
- What data goes into the tool?
- Where is it stored or processed?
- Who can see it?
- How long is it kept?
This is not legal advice. For the final judgement, ask a lawyer. If you first want a look at what your site shows to the outside world, an audit report covers passive, permitted checks only.
Plan tiers and prices
The price depends on scope, not on hours: how many business areas and sessions you need. Prices are excl. VAT.
| Tier | Price | Delivery | Scope |
|---|---|---|---|
| S | $1,200 | 7 business days | 1 business area, 2 sessions |
| M | $1,900 | 10 business days | 3 business areas, 4 sessions |
| L | $3,400 | 15 business days | 6 business areas, 8 sessions, 1 team workshop of 3 hours |
Every tier delivers a PDF plan and a presentation. The plan covers the data and GDPR and KVKK risks and includes a 90-day roadmap. An extra business area costs $450 and an extra session $150. Implementation is not included.
A 90-day skeleton
This is a generic outline. Your plan adapts it to your processes.
Days 1–30: test one process
- Pick one process from the list.
- Map where the data goes.
- Choose a tool based on the plan, not on a demo.
- Test with real but non-sensitive data.
- Keep a person's approval before anything goes outside the company.
Days 31–60: review and correct
- Check a small sample of outputs every week.
- Record errors and fix the instructions.
- Write a short procedure so others can follow it.
- Check the data controls before you expand.
Days 61–90: expand or stop
- Compare the time spent before and after.
- Move to a second process only if the first one clearly paid off.
- If the result is weak, stop and revise.
- Give the decision to one owner, not a committee.
The plan and the build are separate
A plan says what to do. It does not build it. Setting up an agent, connecting systems or building a dashboard is separate work, and owners sometimes expect otherwise.
Anthropic's article Building effective agents draws a useful line. Workflows are systems where LLMs and tools follow predefined code paths. Agents are systems where the LLM directs its own process and tool use. For a first project, a predefined workflow is usually the safer choice. An agent is a bigger step.
After the plan you can order the build from us as its own project, or do it yourselves. You can also stop there, before spending on something expensive. How we work shows the steps.
FAQ
How long does the plan take?
7 business days for S, 10 for M and 15 for L.
What does the fixed price include?
A PDF plan and a presentation, with an opportunity list, data and GDPR and KVKK risks and a 90-day roadmap. Higher tiers cover more business areas and sessions, and L adds a three-hour team workshop.
Is implementation included?
No. The plan stops before any build work, which is quoted separately.
Do I need to pick AI tools before buying?
No. The plan starts from your processes, not from a tool.
Will I get guaranteed results?
No. We do not promise revenue or other outcomes. The plan gives you a controlled order of steps with human checks, because AI output can be wrong.