AI Culture and OperationsAI as part of the work, beyond the tool.

02Who it is for

Does any of this sound familiar?

1. 01The company made AI tools available and ran training, but use remains concentrated in a few people.
2. 02Each team uses AI differently, without knowing which data is allowed, which tools to use or who is responsible for the result.
3. 03People spend time finding information, checking documents, preparing analyses and repeating tasks with known rules.
4. 04Agents work in demonstrations but fail in operations without clear processes, criteria, exceptions and responsibilities.

When leadership asks what changed in the work and what value AI created, the answer needs evidence.

03Deliverables

From isolated use to AI as part of operations.

Both tracks start with the company's actual work. One develops teams' ability to use AI with a method and continuity. The other turns sufficiently understood parts of operations into work that can be delegated to agents.

You can engage either track or combine both, depending on the assessment.

• 01

  To understand where AI can create value

  The assessment shows how teams work, where they use AI, which problems recur and which opportunities merit further work.

  • Assessment of use and maturity by team
  • Activities, bottlenecks and relevant decisions
  • Existing use cases and identified opportunities
  • Initial assessment of impact, frequency, effort, risk and feasibility
  • Available starting measurements to track progress
  • Priorities for development, experimentation or implementation
• 02

  To develop capability within teams

  Learning tracks start with each role's problems and responsibilities, rather than generic tool training.

  • Learning tracks by role, maturity and problem
  • Workshops, mentoring and hackathons on real challenges
  • Experiments: problem, hypothesis, owner, metric and next step
  • Internal materials and reusable practices
  • Use case library
  • Team contacts to sustain ongoing use
• 03

  To make unwritten knowledge part of documented operations

  Before delegating work to an agent, we document with the team how that work actually happens.

  • Responsibilities and limits of each role
  • Inputs, tasks, decisions and outputs
  • Relevant processes and rules
  • Concepts and criteria used by the team
  • Exceptions and escalation paths
  • Required approvals
  • Knowledge needed to carry out the work
• 04

  To put agents to work with clear responsibility

  With sufficient repetition, clear rules and manageable risk, we design the agent to take on only the work that can be delegated.

  • Agent purpose and scope
  • Inputs, context and permitted knowledge
  • Instructions, expected outputs and quality criteria
  • Autonomy, permitted and prohibited actions
  • Exception handling and escalation
  • Human owner
  • Review, decision and approval points

04How the work happens

First we understand the work. Then we decide where AI fits.

The assessment identifies which problems merit further work and which approach fits each one. Not every opportunity needs an agent, and not every agent needs to operate autonomously. You approve the design before implementation.

05Work tracks

Culture to build capability. Agents to change operations.

You can engage either track or combine both, depending on the assessment.

06Working together

AI becomes part of operations with the people who know the work.

07Results

What needs to change in the work.

The result is not the number of licenses, prompts or agents. It is observable change in how work happens, within the approved scope.

• 01

  More repetitive work is delegated while the people making decisions remain accountable.

• 02

  Processes, criteria and knowledge no longer exist only in the heads of the people doing the work.

• 03

  Use cases have an owner, objective, risk assessment, metric and next step.

• 04

  Experiments without sufficient evidence can be reviewed or ended before becoming a permanent cost.

08Governance

An agent does not gain authority just because it becomes more capable.

When needed, we define purpose, owner, permitted data, autonomy, human approvals, quality, monitoring and review.

09Scope and questions

What we agree before starting.