September 29, 2026
We Implemented AI. Now What?
Catherine Toms
AI Expert | Educator | Consultant
The first phase of workplace AI was largely about access and individual adoption.
And that made sense. That’s how the tech itself arrived: a chatbot on your desktop, a blank chat window, you built your own prompts, uploaded your own files and figured out where it could help.
But that model created a new set of problems; limited visibility, duplicated work, inconsistent outputs and AI capability stuck with individuals.
And this “messy middle” is where we’re seeing a lot of organisations stuck right now:
Moving from individual AI use to team-wide adoption needs focus on four things:
1. Process: Redesign the work
Most teams start with: “How can we use AI to improve this process?”
But once we dig into it, we often find there isn’t one clear process to begin with. Different people or teams are doing the same work in different ways. And when you do map it, you often realise the legacy process is no longer fit for purpose: multiple handoffs, duplicated effort and approval steps nobody has questioned because “we’ve always done it this way.”
If the process is already messy, AI will just accelerate more mess. That’s why you need to be willing to go backwards to go forwards.
Instead of asking “how do we make this process faster with AI?”, start with the outcome: What are we actually trying to achieve?
Then map how the work happens today. Where are the handoffs and decision points? Where does work get stuck, repeated or reworked?
Once you can see the workflow clearly, use the 4Ds to redesign it:
DROP
What steps can disappear altogether?
DELEGATE
What can AI take on, with clear human oversight and accountability?
DOUBLE-DOWN
Where should human judgement, creativity or relationship-building increase?
DESIGN
What new or better work becomes possible once capacity is freed up?
2. Context: Build a Shared Intelligence Layer
One of the other risks with the first wave of 1:1 AI is that everyone’s built their own mini ecosystems.
Most teams have hundreds of projects, agents and knowledge bases, each drawing from whatever documents or data people happened to upload.
Which means there’s no single source of truth and a total nightmare for consistency and visibility.
The next phase is cleaning this up: building shared knowledge hubs and a centrally governed contextual intelligence layer.
Think brand guidelines, product information, customer research, SOPs, policies, approved templates, terminology and relevant data sources – clean, organised and ready for AI to work from.
6 steps to building your context layer
Define trusted sources
Decide which business, brand and customer information AI should draw from.
Assign ownership
Make someone responsible for keeping each source current.
Clean and organise
Remove stale, duplicated or conflicting information and keep naming and versions clear. No more v23_FINAL_final.doc.
Connect approved sources
Link AI to the systems, folders and libraries where trusted information lives.
Set access rules
Decide what can be shared broadly and what needs restricted access.
Review regularly
Set a cadence for updating, replacing and retiring content.
3. Governance: Ownership & Accountability
Good AI governance needs to cover both the people using AI and the systems doing the work.
- Human governance
Covers appropriate AI use, quality standards and accountability.
What tasks are appropriate for AI? Where does human review need to happen? Who’s accountable for the final output or decision?
- System governance
Defines what AI is allowed to access and do. As we move towards agents and automations, these boundaries increasingly need to be built into the system itself:
- Permissions and access
- Sandboxing
- Monitoring and audit logs
- Escalation and incident reporting
Move from enterprise policy to task-level guardrails
This is where governance often breaks down. Enterprise policies set the overarching rules, but people are unclear what that means for their actual work.
That’s where we see distributed governance working really well.
- Enterprise policy sets the overarching rules and guardrails.
- Functional leads translate those into practical, task-level guidance for the specific work inside their teams.
Bring the whole team on the journey through skills development, policies and practical guardrails.
4. Culture & People Change
You can give people the tools, training and governance, but if the environment around them doesn’t support behaviour change, adoption will certainly stall.
We look at culture through 4 lenses:
- Safety
Investing time trying things, learning and failing are part of adoption too. People need to feel safe questioning outputs, experimenting and figuring out AI’s limitations. - Space
AI is the highest priority that’s no-one’s job. You can’t ask already stretched teams to embed AI without giving them time to do it. Even 90 minutes a week to apply AI to real work can build new habits. - Social proof
Make AI use visible and normal. Share what people are building, let teams learn from each other and make sure leaders are visibly using it too. - Signals
Look at what gets rewarded and recognised. If you’re asking people to change how they work but still measuring them on the old way, they’ll default to old behaviours.
Where to start: the 90-day plan
FAQs
What are the four foundations of AI adoption?
We look at four things: process, context, governance and culture. How does the work itself need to change? What information should AI be working from? What are the guardrails? And how do you create the environment for people to actually change how they work?
Why does AI adoption stall?
Across most organisations everyone’s rolled out the tools, got board-level ambition and individual tool usage happening, but almost nobody has built the middle layer – the operating model that connects strategy to execution.
How do you scale AI adoption across a team?
Don’t try to do everything at once. Get clear ownership and governance in place, build capability across the team, then pick a few high-value workflows and redesign them properly. Test, learn, prove the value and scale what works.
When do I start?
Successful applicants receive immediate access to our Start Anytime programme, so you can begin learning straight away before joining the live October workshops.
Can you train my team on AI?
Yes. We work with teams to build practical consistent AI capability around the way they work. That means learning how to work smarter with AI, redesigning workflows and building the habits and confidence to use it well.
Do you run AI training for leaders?
Yes. We work with leadership teams on how to drive effective adoption and team-wide change. We cover strategy, governance, capability, workflow redesign and how to lead the change across your team.
Explore the other guides in our AI at Work series, covering the people, strategy and operating changes needed to scale AI across organisations.
Explore more AI resources
Practical guides, frameworks and tools to help you build AI capability, improve how work gets done and scale adoption across your organisation.
