See where your teams get stuck with AI.
Moonford anonymously analyses AI conversations inside your environment to show what fails, what it costs and what to fix first.
See how it worksFindings Ranked by hours lost
Illustrative example
| # | Finding | Team | Owner | Hours lost |
|---|---|---|---|---|
| 01 | The contract archive cannot be reached from Copilot | Sales | Platform team | 290 |
| 02 | The CRM connector returns nothing in one in five requests | Sales | Platform team | 140 |
| 03 | The invoice agent loops until its step limit | Finance | Agent owner | 85 |
| 04 | New joiners wait weeks for a Copilot licence | Operations | IT and identity | 60 |
| 05 | The policy assistant quotes an out-of-date handbook | HR | Application owner | 45 |
The problem
The licences are bought. Now the board wants the cost under control.
Nearly 40% of the time AI saves is lost to rework.
Source: Workday and Hanover Research, 3,200 employees at companies over $100M revenue, November 2025
The signal
The answer is already written down.
Every time someone gets stuck, they say so to the assistant. Almost none of it reaches a ticket queue.
Analysed anonymously, inside your environment. Only the friction is lit.
We analyse friction in your AI layer.
Moonford looks only at friction that affects your AI rollout. Each finding names the team that can fix it.
Connectors that time out, return errors or come back empty, and sources the assistant only partly reaches.
Licences not assigned, missing permissions, connectors not enabled for a team, and repeated sign-in prompts.
Data in your tools that the assistant cannot use: empty fields, duplicate records, answers that take several steps.
Agents that retry in loops, stop at their step limit, pick the wrong tool or send malformed requests.
Requests no connected tool can serve, and connectors that exist but go unused.
More model than the task needs, tokens spent on retries, and the same question answered again across teams.
The output
Every fix has an owner, a before and an after.
- Where work stalls
- By team and by tool.
- What it costs
- In figures your finance team can sign off.
- What to fix first
- Ranked by how many people it gets working again.
410 people in sales could not reach the contract archive from Copilot last month.
290 h
lost last month
1
permission to change
IT
platform team owns the fix
Your data stays where it is.
Moonford anonymously analyses conversations, securely inside your own environment. It reports on AI, tools and work, never on individual people.
Built to work with your AI stack, not to replace it.
Observability tools and AI gateways show how your AI runs. Moonford works alongside them and can use their logs as a further source.
Your AI stack
Alongside it
Assistants and agents
Copilot, ChatGPT Enterprise and your own agents
Conversation historyAI gateway
Routes requests, sets limits, tracks spend per model
Logs, if you run oneAI observability
Traces, latency, errors and tokens per call
Logs, if you run one
Conversation history, and logs if you run them
Where AI work fails, what it wastes and who can fix it.
Output: a ranked list of fixes, each with an owner
Find the waste in your AI rollout.
We are opening a small number of early partnerships with enterprises rolling out AI assistants at scale.
Or email hello@moonford.io