Where to start with AI in your business

Ask your team to spend 10 minutes every evening for two weeks telling what they did, what they repeated by hand and what got in the way. Tasks several people complain about are the first candidates for AI. Pick one process and run a 2–4 week pilot.

In short

  1. 1

    Weeks 1–2

    Each evening every person tells, by voice, how the day went. AI asks follow-ups and groups the same complaints into one summary.

  2. 2

    End of week 2

    A list of tasks several people complain about, with the number of mentions.

  3. 3

    Weeks 3–6

    Hand one process to AI and compare before and after in the same reports.

  4. 4

    Then

    Faster? Keep it and take the next item.

Why it stalls

AI adoption gets stuck at the first step

Owners rarely lack budget for AI. What they lack is a clear view of where AI would bring a measurable benefit in their own company.

Every step-by-step plan starts the same way: set a goal, audit your processes, find where people spend the most time. Good advice — but in a company of 5–50 people nobody has time for that audit. There's no analyst, and the manager sees work by its result, not by how many hours a week go into copying leads from email into a spreadsheet.

A one-off survey

Gives general answers like “too much routine”.

Talking to everyone

Takes the manager a week, and people find it awkward to complain to the boss face to face.

Time tracking and process maps

Done by a consultant — expensive for a small business.

The people doing the work know where the time goes. A short daily account gives what a survey can't: specific tasks and how often they repeat. When three people mention the same manual step two weeks in a row, you see it as a number: six mentions, three people.

Step 1

Two weeks of evening reports

  1. 01

    Invite the team

    People get an invitation and join on their own — no developer needed.

  2. 02

    10 minutes by voice

    Every evening each person tells how the day went. No forms or spreadsheets.

  3. 03

    AI asks follow-ups

    If details are missing, AI asks: “The plan also had a note to the accountant — what happened with it?” It checks the account against plans and tasks.

  4. 04

    A summary for the manager

    A report for each person and a “Complaints & ideas” tab: the same complaints are grouped, with how many times and how many people mentioned them.

What people talk about

  • What they did today and what they didn't get to
  • What took longer than it should have
  • What they did by hand and repeated again
  • What got in the way and what they had to wait for
  • What they would change

Two weeks is the minimum. A couple of days shows one-off problems; two weeks shows what repeats.

How team digitalization works

Step 2

Pick a process for AI from the complaint summary

Not every complaint is a task for AI. Start with a process that meets at least four of these five criteria.

  1. 1

    Several people complain

    One complaint may be a one-off; three are a process.

  2. 2

    Repeats daily or weekly

    Automating a rare task doesn't pay off.

  3. 3

    Easy to check

    A person immediately sees when AI got it wrong.

  4. 4

    Cheap to get wrong

    Keep payments and legal documents away from AI at first.

  5. 5

    Data is already digital

    Leads in email and chats, spreadsheets, messages — not paper.

What usually comes up in reports

“I answer customers' same questions over and over”

An AI front desk in Telegram, by email or on your website AI front desk

“I copy leads from chats and email into a spreadsheet”

Leads from your channels land in the CRM on their own Website widget

“I write scripts and posts”

An AI employee for the role: scriptwriter, news reviewer AI scriptwriter

“I put together a report for my manager”

The report already builds itself from the evening accounts Team digitalization

“Waiting for the client's brief”, “waiting for approval”

That's about how work is organized. AI won't help here, but the complaint is visible and can be solved with a conversation

Some complaints will turn out to be about management — and that's a result too: you see what holds the team back before spending money on automation.

Step 3

A 2–4 week pilot

One process, one team

Launch AI everywhere at once and you can't tell what worked.

The owner is someone who complained

They need the pilot most and will spot mistakes first.

The “before” metric already exists

Two weeks of reports show how long the task took and how often it came up.

AI drafts, a person checks

A customer reply, a lead card, a text — until you trust the quality, a person sends it.

Reports continue

If the complaint disappears from the summary or the task takes less time, the same reports show it.

Step 4

Decide and take the next process

After 2–4 weeks compare the reports before and during the pilot. Faster? Keep it and take the next item on the list. Not faster? Find out what was missing: data, bot setup, or the process wasn't a fit for AI. Take the next process from the same summary — it will have updated by then.

Who owns AI implementation

The owner or a team lead. You don't need a dedicated AI specialist to start: people record reports themselves, AI builds the summary, and the manager decides on the pilot because they own the result.

Mistakes

Mistakes at the start

Buying an AI subscription and calling it implemented

Giving everyone access to a chatbot changes nothing until AI is built into a specific task.

Automating everything at once

A pilot on one process is easier to evaluate and easier to roll back.

Handing AI a messy process

If the rules aren't agreed, AI speeds up the mess too. First agree how the process should run.

Turning reports into surveillance

If an honest report gets someone told off, people stop telling the truth and the summary becomes useless. Reports exist so people can say what gets in their way.

In BotB2B

Two weeks — and you see what to hand to AI

  • People spend 10 minutes telling about their day by voice — AI asks follow-ups and checks against plans and tasks
  • “Complaints & ideas”: the same complaints grouped, with how many people mentioned them and how often
  • Reports for each person, team trends, tasks and CRM — in one account

The Business plan is $49/mo and includes digitalization of 10 team members. Each additional one is $30/mo.

Read next

Questions about the first step

BotB2B / For business

Run two weeks of reports

Sign up and invite your team — in two weeks you'll have a list of processes to hand to AI.