First, the disclaimer: this is a plan, not a results report.
I'm working toward buying a business and haven't run this on one I own. So I won't tell you what happened in my first 100 days. I can tell you how I already use AI when evaluating a business, and how I'd sequence it after a purchase, based on what I do for clients and in my own companies.
I'm writing it as a plan because most advice joins two things that should stay separate: using AI to buy a business and using AI to run one. They're different jobs with different risks. For the buying side in full, see how to buy a small business.
Part one: AI during diligence
AI is useful here. It isn't a reviewer you can trust blindly.
What I use it for:
- Organizing a records pull. Keeping track of what's been requested, received and still missing.
- Reading a long listing book and pulling out every number with its page reference, so I can recompute them.
- Flagging figures that don't reconcile, such as an earnings number that doesn't match its own line items.
- Drafting the questions for the seller, which I then cut down to the ones that change price, closing or a surviving risk.
What I don't let it do: decide anything, or sit between me and a source document.
An AI review of a deal file can spend much of its effort correcting conclusions that had already been replaced, and cite page numbers that don't exist. Give it a corrected second pass and it can still find the most important number in the file. Both are true. The fix is process, not distrust: keep one folder of primary records, mark superseded conclusions as superseded, give the AI every source and ask it to cite where each claim comes from. Then check the citations.
Don't paste tax returns with Social Security numbers into any tool you haven't vetted. Keep personal data out of notes and out of anything you forward.
Part two: before day one
The closing date isn't the start. The first 100 days are decided in the weeks before it.
- Write down how the business really runs. What does the owner do each week that isn't on any document? Interview him. Record it. Turn it into procedures with AI's help, then have someone who does the work edit it. I describe the method in SOPs for a small business.
- Baseline everything. Before you change anything, record what the business does today: response time to new leads, jobs booked per week, missed calls, quote-to-close rate, days to get paid. Without a baseline you can't prove anything worked.
- Check the data you're inheriting. Is the CRM current? Does the phone system log missed calls? Is the analytics tag on every page? A broken measurement system makes every later decision a guess.
Part three: the first 100 days
My sequencing rule: don't touch what already works until you understand why it works. The owner you bought from kept the place running for years, often for reasons that aren't written down. Customers, crews and vendors are all reacting to a change in ownership. That isn't the moment for a rebuild.
Days 1 to 30: listen, document, measure
- Meet the people. Learn the work. Keep the customers.
- Capture processes as they are, not as you'd like them to be.
- Set up the baseline numbers and check them weekly.
- No AI in front of customers yet. The only AI at this stage is internal: notes, transcripts, drafts, summaries.
Days 31 to 60: fix the leak nearest the money
For a service or trades business, the first target is almost always the same: what happens between a lead arriving and a job being booked. Missed calls with no callback. Quotes that go out and never get chased. Past customers nobody asks back.
That is where I'd start, and it's why I put follow-up first when I advise clients. See my rule for what to automate first. Typical early moves:
- Missed-call text-back and a clear after-hours response.
- A consistent follow-up cadence for open quotes.
- A reactivation message to past customers, with consent handled properly. Di-Hy's guide to reactivating old customers covers the method.
- A weekly report with the same few numbers.
Run each one manually first. You can't automate what you haven't validated by hand.
Days 61 to 100: add, measure, keep a human in the loop
- Turn on automation only for steps you've proven manually.
- Give each system a named owner and a five-minute weekly check. Automations die quietly when nobody is assigned to notice.
- Humans review anything that touches money or a customer until the system has earned trust.
- Compare against the baseline on a date you set in advance.
What I would not do
- Replace the people who know the business. The crew and the office manager hold knowledge you can't download. AI should take the repetitive work off them, not make them leave.
- Buy tools before understanding the process. A tool bought in week two will be a tool nobody uses by month six.
- Count hours saved as profit. Freed time only matters if it goes somewhere. Decide in advance what the time is for. I go deeper in why hours saved is not profit.
- Let a chatbot talk to customers before it knows the prices, services and rules. A demo answers well. A live customer finds the gaps.
- Automate a mess. If the team can't explain the process manually, AI will automate the confusion.
Why I'm writing this before I've done it
Partly because it forces discipline. Writing the plan down means I can check later whether I followed it. Partly because the honest version of this topic is thin: most AI-for-business content is written by people selling tools, and most acquisition content ignores AI. If I have a closed deal someday, I'll write what actually happened, including what I got wrong.
If you're looking at a business to sell and you'd like to talk with a buyer who intends to run it and has a plan for the transition, my acquisition page is the place to start. And the reasons these projects fail are in what to check before buying AI.
FAQ
Can AI help with business due diligence?
Yes, for organizing records, reading long documents, flagging numbers that don't reconcile and drafting questions. It needs source documents and checking, because it can state wrong conclusions confidently.
What should you automate first after buying a small business?
Usually lead follow-up: missed-call recovery, quote follow-up and past-customer reactivation. It sits closest to revenue and is easy to measure. Wait until you understand how the business runs before changing anything.
Should you change things right away after buying a business?
Be careful. Customers, staff and vendors are already adjusting to new ownership. Learn why things work, document them, set baselines and then change the parts that are clearly leaking.
Note: This is a plan based on my work with clients and my own companies, not a report from an acquisition I have run. Treat it as method and opinion.