For AI operations leads
Every team wants AI. You're the one who has to deliver it.
Kowalah gives you the pipeline, the build capacity and the evidence: every idea scored, the best ones built on your own Claude account, and outcomes recorded against the KPIs your business already tracks.
The problem
Every team wants to build something. You're tracking all of it.
Requests arrive from every direction, your tracker is out of date the moment you save it, and Friday goes on building the steering update from Slack threads and a spreadsheet. There's a new model or release every week, and you're often the only person doing this.
Work out your number
Hours a week you spend chasing updates and building the steering update × 52 = your hours a year on reporting, before any use case moves
Log one normal week. Count chasing, tracking and reporting separately, because each one has its own skill.
What you're measured on
What you're judged on
The business funds processes that change, with a result it can see. These are the numbers that show it.
- Ideas in the pipeline, scored
- Every team's ideas in one backlog, scored on impact, confidence and ease, so the best ones are obvious.
- Use cases live
- Ideas moved from the backlog to live, with the agent or skill behind each one in the Agent Hub.
- Time from idea to live
- Expert Requests add build capacity, so a shortlisted idea doesn't wait for you to find the time.
- Processes redesigned, with a measured outcome
- The number your sponsor pays for. A process changed, with where it stood before and after recorded.
- Adoption by team
- Usage by person and by team, so you know where to help next.
- Time to produce the steering update
- The update drafted in Claude from the live picture of what's in flight, live and stalled.
The work behind the numbers
The work behind them
Four jobs fill an AI operations lead's week. We start with the one that's costing you the most time.
- 01
Build the pipeline
Collect ideas from every team, map the process behind each one, and score them so the best go first.
Moves: Ideas in the pipeline, scored
- 02
Deliver use cases
Scope the shortlisted ones, set the quality bar, build, and launch to the people who'll use them.
Moves: Use cases live
- 03
Prove the return
Record where each process stands before it changes, record the outcome after, and report it to the people who fund the programme.
Moves: Processes redesigned, with a measured outcome
- 04
Keep track of the estate
Know which AI tools, licences and connectors are in use, approved and up for renewal.
Moves: Adoption by team
Your team, AI-native
The people you work with
You can't run the programme alone. These are the people who make it move, and what changes for each.
Process owners
Today Send ideas and requests by email and Slack, then wait for you.
With AI Map their own processes and raise ideas from their own Claude, straight into your backlog.
Claude Champions
Today Help colleagues when they have time, with no view of who's using what.
With AI Run a champions network you can see, with adoption measured by team.
IT and security
Today Review every tool and connector as a one-off request.
With AI Work from one register of the tools and connectors in use and what each one is approved for.
Your sponsor
Today Gets a status update you built by hand on Friday.
With AI Gets outcomes recorded against the KPIs they own.
What you get
What you get
Each one sits on a job above, and runs inside Claude through the Kowalah Plugin. We set them up for your organisation.
- Skill
Opportunity finder
Build the pipeline
Helps anyone in the business find where AI could help in their own work, and raise it to your backlog with an ICE score.
Runs on
- Kowalah Plugin
- ICE scoring
You get
- Ideas raised by every team, from their own Claude
- Each one scored on impact, confidence and ease
- Each one linked to the process it changes
- Skill
Process diagnostic
Build the pipeline
Shows where a slow process loses its time, and what about it has never been measured.
Runs on
- Kowalah Plugin
You get
- Where the cycle time goes, step by step
- What to measure before you change it
- Skill
Backlog triage
Build the pipeline
Lets you shortlist, park, merge or reject every idea in the backlog, and the person who raised it sees why.
Runs on
- Kowalah Platform
- ICE scoring
You get
- One ranked backlog across every team
- Duplicates merged, with the reasons kept
- A shortlist ready to hand to the build
- Skill
Quality bar
Deliver use cases
Sets the acceptance criteria, gates and monitors for a step before AI runs it.
Runs on
- Kowalah Plugin
You get
- Acceptance criteria for each step
- The gates and monitors that catch problems early
- Skill
Outcomes recorder
Prove the return
Records the value a change delivered against the goal it counts towards, and links it to the agent or skill that produced it.
Runs on
- Kowalah Plugin
- Kowalah Platform
You get
- A baseline recorded before each change
- The outcome recorded after, against the goal
- Credit for the use case that delivered it
- Skill
Programme review
Prove the return
Drafts the steering update from the live operating model, covering what's in flight, what's live, what's stalled and where the gaps are.
Runs on
- Kowalah Plugin
You get
- The steering update drafted in Claude in minutes
- The gaps in coverage, by team and process
- Skill
AI estate register
Keep track of the estate
Keeps a register of the AI tools, licences and connectors in use, what each is approved for and when it renews.
Runs on
- Kowalah Plugin
You get
- Every AI tool and connector in one register
- Shadow AI found before the invoice arrives
- Renewals in view ahead of time
The stack
The stack we bring
Claude at the centre of the programme, with the plugin, platform and frameworks we set up around it.
Claude features your team uses
- Skills
- Packaged instructions that teach Claude how your organisation does a job. Your own, your company's, or ones we build for you. Anthropic
- Projects
- A shared workspace per use case, with the brief and context kept. Anthropic
- Connectors
- Approved connections to your systems, so Claude works with your data where your people already work. Anthropic
Skills and plugins we install
- Kowalah Plugin
- Your operating model, goals and backlog inside Claude, ChatGPT or Microsoft 365 Copilot, with the skills on this page. Free to start. Kowalah
- Kowalah Platform
- The system of record for the programme. Processes, ideas, work in flight and outcomes, in one place. Kowalah
Connectors we set up
- Slack and Microsoft Teams
- Where the Kowalah Agent meets your teams, to raise ideas and get help after go-live.
- Google Workspace and Microsoft 365
- Documents, mail and calendars inside the conversation, seen only as far as the person can already see.
Frameworks and practices we build on
- ICE scoring
- Impact × confidence × ease, out of 1,000, so every idea competes on the same scale. Sean Ellis
- APQC PCF
- A shared map of business processes, so every team describes its work the same way. APQC
- The Kowalah Way
- Phase-gated delivery with named deliverables, and a security checkpoint before any build. Kowalah
Agents
- Managed Agents
- Anthropic's platform for running agents in production. Anthropic
- Agent Hub
- Every agent your organisation runs, with its owner and its daily cost, on one screen. Kowalah
- Kowalah Agent
- In Teams, Slack and Google Chat, helping people raise ideas, get help and keep using what's been built. Kowalah
One process, start to finish
Before and after: the Friday steering update
From chasing every team for an update to a draft you check and send.
Before
- Ask each team lead for an update on Slack.
- Copy the answers into the tracker.
- Chase the ones who didn't reply.
- Guess at outcomes nobody recorded.
- Build the slides on Friday afternoon.
With Claude
- Ask Claude where the programme stands.
- It drafts the update from the live backlog and what's in flight.
- You check it and send it.
The Kowalah Way
How we do it
- You're not doing this alone. A named Client Partner and Expert Requests give you a team behind you, so the backlog moves without you building everything yourself.
- A method to follow. Ideas, scoring, build and outcomes run the same way across every team, so you're following a method rather than inventing one.
- Proof the business recognises. Every use case is tied to a KPI a business owner already tracks, so your update reports what changed for them.
How we'd work with you
How we'd work with you
Start with an AI health check to see what's in place, or take all four together and we run the programme alongside you, from the backlog to the outcomes. Every pillar is tied to a number your business already tracks.
Strategy & Advisory
What should we do?
-
AI health check
What you've got today and whether it's set up right: accounts, connectors, agents and skills, which connectors to approve, and what to fix first.
-
AI programme plan
The goals to measure, the backlog scored, the processes to change first, and a 12-month plan your sponsor can read.
Design & Build
Build it for us
-
Quick start
Your organisation set up and working in Claude: accounts and permissions sorted, approved connectors in place, and your first skills built.
-
Use case build
Shortlisted ideas built as agents and skills on your own Claude account, each defined like a hire with the KPI it moves.
Change Enablement
Get my team using it
-
Claude Champions
A champions network inside each team, briefed and supported, with adoption measured.
-
Team training
Training by role, built around the work each team does every week.
Managed Services
Keep it improving
-
Expert Requests
New skills, agent changes, fixes and training on request. Bought as a pack, or quarterly or annually.
-
Kowalah Platform
Your backlog, agents and outcomes in one place, with new accelerators added as we build them.
Questions
What AI operations leads ask first
What does an AI operations lead do?
They run the AI programme day to day: collecting ideas from every team, deciding what gets built first, getting use cases live, and reporting the results to the people who fund the programme.
Will this replace my role?
No. You decide what gets built and in what order. We add build capacity and a method, and the outcomes are credited to the use cases you delivered.
What if some teams use ChatGPT or Copilot?
The Kowalah Plugin is available in Claude, ChatGPT and Microsoft 365 Copilot, so teams can raise ideas from the assistant they use. The agents we build run on Claude, on your own account.
Do I need my CMO or CRO on board first?
It helps. Your sponsor owns the number the programme is meant to move, so a plan tied to their KPIs is the quickest way to get the budget behind the next use case.
Ready when you are
Bring your backlog. Leave with a plan.
A free 45-minute working session with a Kowalah consultant. Bring the ideas and requests you're juggling; we map them to the processes and goals behind them, and you leave with a one-page plan: the first three use cases to deliver, how to measure each, and what to report.