01Service · Culture
AI Culture and OperationsThe whole company working with AI.
We take AI into the work of every area and every level, starting from the company's real challenges, until people build their own solutions. And we write down with every area how the company works, so the repetitive work of each role can go to an agent while the decision stays with whoever approves.
What arrives
- A plan of tracks per area and per level
- An AI usage policy per area
- Agents running in each role's work, with the human decision kept where it is required
How it is boughtOngoing culture cycles, and a role-by-role agent project.
02Who it is for
Does any of this sound familiar?
- The company rolled out the tool, ran a training session, and AI stayed with the people who already knew how to use it.
- Every area uses AI its own way, and nobody knows which data is going into which tool.
- Every area repeats by hand the work that needs nobody's decision: answering the same question, copying data from one system to another, checking the same document.
- When someone tries to automate it, the agent gets wrong what anyone inside would get right, because nobody wrote down how the company works.
When the board asks what the company got for the investment, the person who answers is you.
03At the end
Where the company stands when the work is done.
People in every area use AI in the week's work, not just those who already knew how.
Every area knows which tool to use, for what, and which data never goes into it.
Each role's repetitive work runs with an agent, and the decision stays with whoever approves.
04Instead of
What usually gets tried first, and where it stops.
Rolling out the tool and waiting for adoption
Where it stops
Use stays with whoever already knew how.
One general training for the whole company
Where it stops
It teaches the tool, not each role's work, and the habit is gone the following week.
Letting each area automate on its own
Where it stops
One agent per person appears, with no owner, no limit and no usage policy.
05In your market
The same problem, under the name it has in your sector.
Real Estate
Collections, credit analysis and investor relations each use AI their own way, and nobody knows whether a buyer's contract went into a tool the policy does not provide for. Checking which portfolio base is right repeats every week; deciding the credit stays with whoever approves.
Non-profit Sector
Fundraising and marketing use AI on different campaigns, and what donors said in the survey still sits in a file nobody reads. Checking a recurring donor's declined charge repeats every month; deciding the reactivation campaign stays with whoever approves.
Franchises and Multi-Unit Retail
Operations, expansion and controllership use AI in different ways, and each unit's spreadsheet goes into whichever tool each person chose. Checking the cash-up store by store repeats every day; deciding a franchisee audit stays with whoever approves.
06The proposal
Two ways to engage, and they fit together.
One module looks after people and habit; the other, after the work of each role. You can start with either.
Engaging both
A solution born in a workshop or a hackathon leaves with an owner, a next step and a destination. When the destination is an agent for a role, it is built in the agent module.
07Module 1
AI Culture Across the Company
For when AI stayed with a few people and every area uses it differently.
A diagnostic of a few weeks opens the first cycle. Then three-month cycles with one-to-one mentoring for whoever decides, workshops, facilitations and hackathons on the company's own challenges, and periodic online training.
Delivery
What stays with you
Before the first event
- A diagnostic of AI use per area and per level, built from a survey and interviews
- A map of each area's real challenges, prioritised for the events
- A plan of tracks per area and per level, with the format of each stage
See the rest
- An AI usage policy per area: which tool each person uses, for what, and what never goes into it
At every event
- A facilitation script per event, with the challenges and the material
- The solutions built in workshops and hackathons, each with a named owner, a recorded next step and a destination: an agent for a role, an application or a product
For whoever decides
- A decision record per cycle, with the alternatives ruled out
- A written criterion for retiring an initiative that is not holding up
- A technical reading of vendor proposals, in writing
For the use to last
- The online training material for each track, which stays with the team
- A library of prompts and use cases per area, kept by the focal points
- A network of focal points per area: your people, prepared to sustain the use between events
See the rest
- An adoption report per cycle, per area and per level
Stages
How it works, stage by stage
Stage 1: Research and diagnostic
We measure how each area and each level uses AI today, interview the people who do the work and collect the real challenges that become the events' material.
Done when every area and level has a picture of today's use and its real challenges collected.
Stage 2: Plan of tracks
Each track gets a format and a usage policy. The board approves the plan before the first event.
Done when the board approves the plan of tracks and the usage policy.
Nothing moves on without your approval.
Stage 3: Mentoring, workshops and hackathons
One-to-one mentoring for whoever decides; workshops, facilitations and hackathons on the area's real challenges. Your team builds; we prepare, facilitate, and fail what does not meet the track's criterion.
Done when the solutions built at the event leave with an owner, a next step and a destination.
Stage 4: Between events
Periodic online training on the track's own use cases, starting from what each person built since the last session.
Done when each person takes the track's use case into the week's work.
Stage 5: Measurement and the next cycle
We measure adoption per area and per level, and the next cycle starts where it did not take.
Done when adoption is measured per area and level, and the next cycle has its starting point.
Participation
Who from your team takes part
- A sponsor on the board
- One focal point per area
- The people of each level, in the events of their own track
Measurement
How you know it worked
Measured in your own operation, before and after.
- People using AI in the week's work, per area and per level, before and after
- Areas with nobody people turn to for AI
- Solutions people built themselves that are still in use after the event that produced them
- Challenges taken to a hackathon that reach day-to-day work
- People who took customer data to a tool the policy does not provide for
- AI decisions brought to a cycle and closed inside it
- Initiatives retired for failing their own stated criterion
- Decision records a successor can read without asking anyone
08Module 2
Role-by-Role Agent Rollout
For when the repetitive work of each role is already in plain sight and the AI does not know how the company works.
A project in five stages, from the audit of each role's work to the handover to the area, with the agent team's design approved before anything is built.
Delivery
What stays with you
The work, sorted
- An inventory of each role's work, separating repetition from decision
The operation, written down
- A company glossary, one definition per term, read the same way by people and by agents
- A step-by-step drawing of the critical processes, with the point where each one waits for a human decision (BPMN)
- A matrix per process of who executes, who approves, who is consulted and who is informed (RACI)
See the rest
- A context document the agent of each role reads before it acts
The agents
- The agent team's design, with what each one reads, the tools it may use and the point where it hands back to the person
- Agents running in each role's day-to-day work, with the human decision kept where it is required
- The limit that blocks an action the decision map reserves to a person and hands it back to them (a guardrail)
The proof and the handover
- A before-and-after report, with the indicators measured in your own operation
- An operating manual per agent, with a named owner in the area (runbook)
Stages
How it works, stage by stage
Stage 1: Audit per role
In each role we separate the task that repeats from the one that needs a human decision, working from interviews and day-to-day documents.
Done when each role has an inventory of what repeats and what needs a decision.
Stage 2: The operation written down
We write down with every area how the company operates: the glossary, the critical processes step by step, and who executes, approves, is consulted and is informed.
Done when every critical process is written down, with who executes and who approves.
Stage 3: Agent team design
What is only executed can go to an agent; what is approved stays with a person. The board approves the design before anything is built.
Done when the board approves the agent team's design.
Nothing moves on without your approval.
Stage 4: Building with the person in the role
Each agent is built with whoever holds that role, on their real work.
Done when each agent runs in the role's work, with the person who holds it.
Technical detail
Each agent declares what it must read, the tools it may use and the point where it hands back to a person.
Stage 5: Measurement and handover
We measure again, in your own operation, what we measured in the audit, and each agent passes to an owner in the area.
Done when each agent has an owner in the area and the before and after is measured.
No figure of ours enters that arithmetic, which is what makes the result yours to argue with upstairs.
Participation
Who from your team takes part
- A sponsor
- One focal point per area for the mapping
- One person from each audited role, building alongside us
Measurement
How you know it worked
Measured in your own operation, before and after.
- Weekly hours each role spends on repetitive tasks, before and after
- Repetitive tasks on the inventory now done by an agent, over the inventory's total
- Cases an agent hands back to a person, and how many of them were actually a decision
- Decisions reserved to a person that an agent executed without handing back
- Terms two areas define differently, before and after
- Agents adjusted by the area itself after the handover
09After the handover
What keeps running without us.
The culture cycles go on for as long as adoption calls for them, and each one starts where adoption did not take in the last. Each agent stays with an owner in the area and an operating manual, and each area's focal point keeps the library of use cases.
10Questions
Common questions
Do you do the work for us?
Your team builds. In the workshops we prepare, facilitate, and fail what does not meet the track's criterion. With the agents, each one is built together with whoever holds that role, and passes to an owner in the area.
Will the agent decide instead of people?
The agent takes the task that is only executed. What someone approves stays with that person, and a written limit blocks the action and hands it back to them. What belongs to whom is decided by each process's matrix, drawn up with the area itself.
What about our customers' data in AI tools?
Everything runs inside a usage policy per area, which says which tool each person uses, for what, and what never goes into it. We measure how many people took customer data to a tool the policy does not provide for.
Where do we start: culture or agents?
If AI stayed with a few people, culture comes first. If a role's repetitive work is already in plain sight, the agents can come first. Workshop solutions that become an agent move on to the agent module.
11Other services
The other services we offer
12Contact
Show us where AI stalled
Tell us which teams already use AI and where it stopped at the tool. Giulliano Soares is who answers.
- Speak with the microphone on
- Type whenever you prefer
- The contact is recorded at the end
Where would you like to start?
Type, or tap Talk by voice.