MMOOZZGG.com

AI assistant

In short

Mike Mozg, technical director and chief AI architect at Devblock, describes how a 120-engineer IT company became an AI development center between 2020 and 2026. The change rebuilt how the team works rather than replacing people. The main takeaway: AI needs written work rules, so start with a repeating process, test it in a pilot and then scale.

Key ideas

  • Devblock grew from a classic IT company of 120 engineers into an AI development center between 2020 and 2026.
  • The approach: pick processes where work repeats, give agents access to working systems, write the rules, test in a pilot, then scale.
  • The transformation rebuilt how the team does its work; it did not replace people.
  • AI does not fix a process that does not exist: if work rules are not written down, the agent has nothing to follow.
  • The leader chooses the first process and agrees on the metric before the start; at Devblock the technical director led the change.

Prepared by AI from the article

Engineering team transformation: how 120 Devblock engineers became an AI team

AnalysisIT

Medium4 min read

From 2020 to 2026, Mike Mozg was technical director and chief AI architect at Devblock. In that time, he turned an IT company of 120 people into an AI development center. This is an engineering team transformation seen from inside: not a replacement of people, but a rebuild of how the team does its work.

Engineering team transformation: how 120 Devblock engineers became an AI team
On this page · 7 sections

What it was and what it became

It was a classic IT company: 120 engineers, client projects, and a speed limited by the number of people. After the engineering team transformation, it became a company where engineers work together with AI agents that write, check and release code around the clock.

What approach stood behind the transformation

The approach Mike now uses with clients of MMOOZZGG.com: start with processes where work repeats, give agents access to working systems, write down the rules, test in a pilot, then scale. Models are chosen for each task by quality and price, not one model for everything. Data and access stay under the company’s control.

What this means for a company outside IT

You do not have 120 engineers, but the mechanics are the same. You have people who spend the day on repeated work: sorting requests, answering customers, reports. Engineering team transformation starts not with buying subscriptions but with a review of processes and one pilot with a metric. Then an AI department of 3–5 agents is built in 4–6 weeks, with the price and term fixed in the contract, and staff move to work that needs a person.

What changes in management when agents do part of the work is covered separately: how to manage a team where people and AI agents work together.

Numbers worth measuring during the transformation

Engineering team transformation: how 120 Devblock engineers became an AI team

Speed: how much time passes from a task to a result. Cost: what a unit of work costs with agents and without. Quality: the share of errors and the share of cases passed to people. Your company will have its own numbers, but measure the same three things from the first day of the pilot.

What the leader and the team need to give

From the leader: choose the first process and agree on the metric before the start. From the team: one owner per process and half an hour a day for disputed cases. From the company: access to working systems in the first three days and a month of sample conversations. Training staff to work with agents is part of building the department.

What the transformation does not give

It does not fix a process that does not exist: if the work rules are not written down, the agent has nothing to follow. It does not take responsibility off the leader: agents prepare and carry out, a person makes decisions. And it does not happen in one day: a pilot takes 14 days, a department 4–6 weeks, and the next process is chosen by the numbers of the previous one.

Who leads the transformation

At Devblock, the technical director led it. Small and mid-sized businesses usually have no such person. That is what the AI director format is for: Mike spends 5 hours a week running the quarterly AI plan, the AI department, contractors and the IT budget, with a fixed monthly fee in the contract, for no more than 3 companies at a time.

Questions and answers

What is an AI transformation of a company?

A rebuild of how the company does its work: repeated tasks move to AI agents in working systems, people stay on decisions and hard cases, and the result is measured with metrics.

Do you need to lay off staff?

That is not the goal. At Devblock, the team became an AI team. In a small business short of staff, an agent more often fills an open vacancy than replaces a person.

Where do you start the transformation?

With a 30-minute process review and a pilot on one process: 14 days, the price fixed in the contract, the metric in the contract.

How do you move a team to AI without stopping current work?

One process at a time. The pilot runs for 14 days next to normal work: the agent takes the flow, staff sort out disputed cases. The other processes do not change until the first one gives a number.

How is AI adoption in an IT company different from adoption in retail and services?

The processes differ, the order is the same. In an IT company, agents write, check and release code. In retail and services, they take requests, book customers and prepare reports. In both cases, first the rule and a pilot with a metric, then scale.

Do you need your own AI development team to start?

No. A contractor builds and connects the agents, and you keep the system, a perpetual license, the data and the access. You can buy out the source code separately.

Publications

Contact us