How Do We Start With AI in the Business, What’s the First Step?
The most common answer to how to start with AI is wrong: “by subscribing to a tool”. The right answer is dull but saving — with a one-week hour hunt. 🪜
Short answer: four stages — hour hunt, data and rules, pilot, rollout and measurement. The tool arrives at stage three; trials starting at stage one get abandoned.
Below: the four stages, week one’s work, the pilot logic and a typical timeline. 🗓️
Stage one: the hour hunt
The real answer to how to start with AI lives here.
Stage two: data and rules
The most valuable preparation before setup.
Stage three: the pilot
The tool can finally enter.
One job, small team, human approval
A template gets written for the chosen job, tried with two or three people, and every output gets checked. Problems get solved while small, and the team learns what the tool can and can’t do. That learning makes picking the next jobs easier. 👥
Stage four: rollout and measurement
Where the gain becomes permanent.
What’s a typical timeline?
Writing the expectation down saves the project.
First week, first month, first quarter
Week 1: hour hunt and job selection. Week 2: minimum data set and rule writing. Weeks 3-4: template writing and pilot use. Month 2: measurement, correction, rollout. Month 3: the second job. By the end of month three you talk in numbers; timing sits in the results article, all questions on the consulting page. 🗓️
📝 Field Notes
In a team that took the hour hunt for surveillance, the records came back empty. Once we stated the aim plainly and put our own work in the same table, it filled — and the top row was a job nobody expected. The obstacle to measuring isn’t technical; it’s about trust. 📋
📖 Quick Glossary
Hour hunt: listing repeating work with its time cost. Minimum data set: the smallest information package needed to start. Template: a reusable instruction and text skeleton. Pilot: a first implementation on one job.
⚡ Quick Summary
Four stages: hour hunt, data and rules, pilot, measurement. 🪜 The tool arrives at stage three. Rules fit on one page. One job, small team, human approval. The second job waits until the first template settles.
🎯 Next Step
Let’s do week one together: the hour hunt and job selection are free — the digital audit. Scope on the consulting page. 🗓️
Frequently Asked Questions
Sık Sorulan Sorular
Four columns suffice: job name, times per week, minutes each time, does it run on writing. Everyone records their own work; say plainly that the aim is diagnosis, not supervision. The filter logic sits in the which-jobs article. 📋
The one with the highest weekly total that is text-heavy. It usually turns out to be quote preparation, customer correspondence or report assembly. One row gets picked — not two, and certainly not three. 🎯
Three things suffice: current price and scope, frequent questions with answers, two or three well-written examples. These gather in a day and determine most of the output quality — detail in the output article. 📁
Three lines: what may be entered, what never may be, who approves the output. Those three lines don’t exceed a page and form the most critical part of the setup. Rules, not bans — bans produce hidden use. 🔒
Because asking the same thing differently every time produces different results. A good template — what’s wanted, which information gets supplied, in what tone — makes output repeatable. Without a template, the gain is accidental too. 🧩
At the start and in week four. The same job’s before-and-after duration gets compared. An unmeasured gain goes unnoticed and support erodes — the method sits in the return article. 📊
Once the first template has settled and the team has adopted the rule. Changing three jobs at once splits attention and none becomes a habit. Sequencing beats simultaneity. 🚦
A rough one-week record takes a few minutes a day and its aim is diagnosis. Saying so plainly raises participation noticeably. A job picked without measurement usually starts from the wrong place.
You can, but a subscription bought without knowing which job it solves becomes unused by month three. Trying it out is useful; make the company decision after the hour hunt. The sequence isn’t lost speed, it’s insurance.
The root of resistance is usually the fear of “will I lose my job”. State the aim plainly: shortening the repetitions nobody enjoys. Putting the most resistant person on the pilot team is often the fastest fix.
Source: PMI — project management library