InsightsExecutive briefing

What leaders need to decide before everyone gets an AI colleague.

A briefing for leadership teams on what changes when every person has AI they can direct, connected to the systems they already use, and what to do first.

Tim Bond
Tim BondFounder, CogniScale · September 2026 · 5 min read
THE IDEA IN ONE LINE

Give people AI colleagues to think with and delegate work to, through approved connections.

In brief

This CogniScale briefing sets out what leadership teams need to decide before giving every person an AI colleague connected to their work and systems. It covers the operating model, control, ownership and first steps.

The moment we are in

The opportunity is accelerating faster than most leadership teams realise.

Most organisations have given their people an AI chat tool, and that was the first step. The latest frontier models can now work across the systems a company already uses: reading the inbox and the diary, preparing documents, updating records and following work through to completion. AI has moved from answering questions to carrying out work.

The biggest mistake I see is a lack of education. Unless people are experimenting and learning, and that includes the chief executive and the finance director, an organisation falls behind, and it falls behind quickly.

What I have learned in my own companies

We run our own companies this way first. We built our platform for our own use, connected it to every system we work in, and our AI colleagues act on our behalf within the rules we set. For clients, we begin by agreeing access and review arrangements, including which actions need approval and which controls are available. With agentic AI connected to our full technology stack, a small number of us now do the work of eight or nine people.

These examples come from my own working setup. My team principal prepares me for meetings, drafts my emails, monitors my inboxes, manages my diary, analyses call transcripts and summarises meetings. It has sent invoices, helped me plan strategy, managed a legal matter, updated our website, built a microsite, produced presentations and built client surveys. The list keeps growing.

What most leaders have not yet seen is what happens when you connect your systems and give every person AI they can direct. The agency that follows changes how a company runs.

Three ways to work with AI

1The team principalEach person's partner across the whole working day; for a chief executive, it works like a co-CEO. It knows their goals, inbox, diary and meetings, advises them and carries out work on their behalf.
2Connected ColleaguesSet up for a specific task that recurs, such as the month-end pack, a weekly report or the first draft of a proposal. They remember what they have learned from one session to the next.
3AutomationsTasks that run on a schedule, overnight or every few minutes: filing and summarising meeting transcripts, checking inboxes for anything that needs a person, preparing the day's priorities before the morning.

Organisation-wide intelligence is coming soon. An optional Workspace Brain will keep agreed knowledge current for its members, so their race teams can see what has changed, what has been decided and what has been promised.

Staying in control

Clear instructions, appropriate access and human review work together.

The question I am asked most often is whether it is safe to give AI this much access. A team principal works with the same access its person already holds, and within a constitution: the written rules that set out how it works and what it may do. A written brief sets expectations; permissions and checks provide additional control.

Your IT team approves every connection before anything is switched on. Each person directs their own team principal, in a conversation they are part of, and their private conversations do not become shared knowledge.

The platform itself is built with AI. Each change is written by one AI model, reviewed by a second, independent model, and tested and security-checked automatically before it goes live. Our Trust Centre sets out, for your IT team, exactly what is in place today.

What leaders should do first

Leaders cannot delegate their own understanding. My advice to any chief executive who wants to be closer to this is to start personally, this month:

  • Listen to some of the many good podcasts on AI.
  • Test drive the Formula AI Control Centre by plugging in your Claude licence.
  • Set up voice-to-text, such as Whisper, so you can hold down a key and talk your thinking through with your AI, as you would with a colleague.
  • Set up a task that runs on its own, such as a summary of your inbox waiting for you each morning.
  • Keep experimenting, and expect your leadership team to do the same.

The endorsement of the chief executive, the finance director or both is what moves an organisation. Once the leadership understands what is possible, the rest follows.

The journey

Every CogniScale client follows the same structured journey, beginning with one team and extending once the value is evident.

Prove it with one team
  1. A leadership conversation to agree the priority to address first: capacity, cost or growth without stretching your people.
  2. Surveys and the AI Opportunity Report, showing where time goes and where to begin.
  3. A systems and security review with your IT team.
  4. Connecting the systems your IT team has approved, with each person adding their own AI licence (today, Claude from Anthropic).
  5. The first team goes live, each person with an AI race team and the first Connected Colleagues.
Roll it out
  1. PLAN · BUILD · SCALE: every team learns the platform and builds its own colleagues.
Organisational intelligence
  1. A Workspace Brain for shared organisational knowledge (coming soon).

Cost and ownership

You bring your own AI licence (today, Claude from Anthropic) and pay your AI provider directly. We provide the platform, the set-up, the training and the support. Your organisation retains its roles, knowledge, workflows and permissions, and no single model provider becomes the operating system for its work.

Evidence

Our method began inside our own companies, and it has been proven with client teams through our PLAN · BUILD · SCALE programme.

TechPros, our own business75%lower content production costs
SimCorp, through our programme86%less time across four pilot content tasks: 40.5 hours of work completed in 5.5
BTG, through our programme14+agents built by one marketing team in six weeks

The human outcome

What an organisation does with the time released depends on its business and its market. Some give people time back. Some take on more work with the same team, or grow without stretching their people. And some will need fewer people: where routine tasks can be fully automated, certain roles become redundant. Leaders should say so plainly and plan for it.

As AI use grows, leaders need to plan how responsibilities change and where people's judgement, creativity and relationships matter most.

The short version

Connect your systems. Give every person an AI race team they manage, and let each team build the colleagues its work requires. Learn first-hand, and extend on the evidence.

Start with a clear picture.

Short surveys of your leaders and team show where time goes and where the pressure sits. Your AI Opportunity Report then recommends where an AI race team could help first, and how it could change what it feels like to work. No platform commitment is required.