MMOOZZGG.com

AI use case map

How to implement AI in business with a map: 7 steps from review to plan

6 min read3 sources

How to implement AI in business so that the result is measured in money, not by the fact of a launch: first build an AI use case map, then implement from it. Below are the 7 steps we use to build it, each with three lines: what we do, what the company gives and what comes out. The first step, a 30-minute review, is free.

On this page · 8 sections

Step 1. A 30-minute review

We do: in a 30-minute talk we learn what drives the company, where it hurts most and which process makes sense to start with. We say directly whether you need a full map or one solution is enough.

The company gives: the owner or the chief executive on the call and a short story about the business: departments, systems, what you have already tried with AI.

The output: a clear next step. The review is free and commits you to nothing.

Step 2. AI due diligence

We do: AI due diligence, a full check of the company with one question: where AI will bring money and where it will bring risk. We come to the company, interview department heads and review processes, data, systems and costs.

The company gives: department heads’ time for interviews, exports from the CRM and the accounting system, a list of subscriptions, contractors and vendor contracts.

The output: a description of how the company really works, not how the rules say it works. According to BCG, about 70% of difficulties in AI implementation relate to people and processes, 20% to technology and only 10% to algorithms, so we start with people and processes.

Step 3. A loss map

We do: we list the processes where the company loses time and money and put a money value on each: staff hours, salaries with taxes, errors and rework.

The company gives: headcount and payroll by department, and at least rough work volumes: leads, calls, documents, reconciliations.

The output: a table of losses by department, sorted by amount. It shows where AI can return money and where the gain is too small to bother.

Processes with small losses stay off the map, even if a fashionable solution exists for them. This way the budget is not spent on small things.

Step 4. Choosing solutions on the market

We do: for each process we check the market: ready services, AI agents, local models. We suggest custom development only where nothing ready exists.

The company gives: the requirements that cannot be broken: which information must not go outside, which systems the solution must work with and who will use it.

The output: several options for each process, with prices and differences. Large businesses choose the same way: in a survey of 150 CTOs by a large Russian tech company and Yakov and Partners, 78% use ready AI solutions from outside partners, and 71% fine-tune ready models instead of a full training cycle.

Step 5. Calculating the effect

We do: we calculate savings, growth, investment and the payback period for each solution. We take salaries with taxes, subtract support costs and estimate the AI share of work by process, not by job title.

The company gives: confirmation of the numbers from the loss map and the goal by which you will judge the result: response time to customers, the number of processed leads, department costs.

The output: the effect in money for each solution and a metric you can write into the contract and check after launch.

Step 6. Risks and data

We do: we decide what can go to AI and what cannot, where to store data under 152-FZ, Russia’s personal data law, and how not to depend on one vendor. We also check what happens when AI makes an error and who will notice it.

The company gives: a lawyer or a security officer, a list of systems that hold customers’ personal data and the current vendor contracts.

The output: a list of risks for each solution and a rule for each: allowed, banned or allowed with a condition.

Step 7. The plan and owners

We do: we put together the rollout order and the budget, assign an owner to each solution and decide which staff need to learn what.

The company gives: the chief executive’s decision on who will join the AI department: process owners, the IT department, a lawyer and the person in charge of training.

The output: an AI use case map with calculations in money, a step-by-step rollout plan and an answer to whether you need an AI department and what kind.

If the company has no AI department, the plan says who it starts with. In a small business, that is one owner and an AI director for a few hours a week.

What you get at each stage
  1. Stage 1

    AI due diligence

    We study the company from inside: where AI is needed and where it is not

    • Interviews with heads of key functions
    • Process map: from order to payment, from raw materials to shipment
    • Audit of data and its quality
    • Systems review: 1C accounting, ERP, CRM, MES, document flow
    • Spending on staff and contractors
    • AI tools and subscriptions already bought
  2. Stage 2

    Effect calculation

    We put a money figure on what a process costs now and what AI will save

    • Cost of each process: hours, payroll, errors, downtime
    • Choice of ready-made solutions on the market
    • Savings and growth in money for each initiative
    • Cost of rollout and ownership
    • Payback period, NPV and IRR
    • Priorities: effect against complexity
  3. Stage 3

    Digital modeling

    We run scenarios on a model before any money is spent

    • Financial model of the company with its AI initiatives
    • Three scenarios: base, upside, stress
    • Sensitivity to key assumptions
    • Impact on EBITDA and cash flow
    • Where roles change and hours free up
    • Risks: data, Russia’s personal data law (152-FZ), suppliers
  4. Stage 4

    Consultation

    We hand over the roadmap and the plan: what, in what order, at what cost

    • AI roadmap: what, where, with which tools and at what cost
    • 12-month plan with a budget
    • Design of the AI department: roles and owners
    • Technical brief for contractors
    • Rules for staff on working with AI and data
    • Defense of the results before the board

Next, if you wish: Rollout and running of the AI department under a separate contract

How long it takes

The whole map takes 7 weeks. The first week goes to preparation: the contract and the brief, a data room, the data request and calls with management. We start work after the contract is signed.

Then come six weeks of work. AI due diligence of all key departments runs for three weeks, then a week for the effect calculation, a week for digital modeling and a week for consulting and presenting the results to the board.

Steps 1–7 above fit these stages: the review and losses go into due diligence, choosing solutions and the effect into the calculation, risks into modeling, and the plan and owners into consulting.

Project timeline

7 weeksfrom $93,000

  1. 1 week Preparation

    Contract and brief, data room, data request, calls with management

  2. 3 weeks AI due diligence of all key departments

    Starts once the contract is signed

  3. 3 weeks Calculation, modeling, consultation

    Result: an AI roadmap with the effect in money

  • Stage 1

    AI due diligence

    3 weeks

  • Stage 2

    Effect calculation

    1 week

  • Stage 3

    Digital modeling

    1 week

  • Stage 4

    Consultation

    1 week

1 week of preparation and 6 weeks of work. The final price is fixed after the NDA and the scope are agreed.

Questions and answers

What does the company need to give for the map?

A decision by the chief executive and time from department heads for interviews. You also need exports from the CRM and the accounting system, headcount and payroll by department, and a list of subscriptions and vendor contracts. According to BCG, about 70% of AI difficulties relate to people and processes, so the talk with department heads matters more than any export.

How long does the map take?

7 weeks: one week of preparation and six weeks of work. AI due diligence takes three weeks; the effect calculation, digital modeling and consulting take a week each. We fix the term in the contract together with the price.

Sources and fact-checking

Checked by the editors on 14 October 2026.

  1. Yandex and Yakov and Partners: a study of AI in Russia, 8 December 2025 (yandex.ru)
  2. BCG: 74% of companies get no tangible value from AI, 24 October 2024 (bcg.com)
  3. Almanac “AI in Industry”, MIPT AI Institute, 2026 (aireport.ru)

Found a mistake? Tell us

Publications

Mike Mozg

Agentic AI use cases at AT&T: nearly 500 processes now run on agents

AT&T no longer uses AI only to help staff work faster: it is rebuilding whole business processes around agents. SiliconANGLE reported on 14 November 2026 that nearly 500 agentic AI use cases at AT&T are already in production. Here is how the company rebuilds them, where a human still checks the agent, and what other large companies can take from it.

News4 min
Contact us