Discipline
Automation and agents
Building what works on its own and keeps working when nobody is watching it.
8 services in this discipline
This is yours if
Three signs that this is yours.
- Hiring is frozen and volume keeps going up.
- Last year's pilot is still a pilot.
- A process that matters depends today on a shared spreadsheet.
Use this today, without hiring anyone
From use case to production
The five reasons a pilot does not cross over, in the order they show up. Mark the ones that are yours before calling anyone: if it is three or more, the problem is not the model and you already know it on your own.
- There is no owner. Nobody has it in their job description that this works on day two hundred. It is the number one reason and the cheapest to fix.
- The integration was never really done. The pilot ran on data exported by hand. Production means connecting to what already exists, through the paths that are already supported.
- The real process is not the documented one. The pilot was tested against the manual; the operation works another way, with exceptions nobody wrote down.
- The cost per unit was never calculated. In a pilot, consumption is invisible; in production it is a monthly invoice somebody has to approve.
- There is no shutdown criterion. Without the rule for when it gets retired, nobody dares switch it on for real.
All five come down to one. Name the person. If you cannot say out loud who answers for this six months from now, the pilot will not cross over.
It comes from From use case to production. Every service in this discipline brings its own.
The 8 services in this discipline
How many hours a month go into work that could simply not exist?
They gave me a productivity target without any extra headcount.
On the data from your own tickets: how many hours go into repetitive work, which ones get eliminated and in what order.
See the situationFrom use case to productionWhy is your AI pilot not making it into production?
The pilot worked, everyone applauded and eight months later it is still a pilot.
From demo to real operation: with an owner, an operating manual, monitoring, cost per outcome and a shutdown criterion written before it gets switched on.
See the situationCustom operational applicationWhat do you do when a process that matters lives in a spreadsheet?
We run it in a spreadsheet and no product on the market does exactly this.
A small, specific application for your process, with an operating manual, a named owner and the three year cost of operation stated in the proposal.
See the situationDomain document assistantHow does your team find what the organization already knows?
Everything we know is written down and nobody finds it when they need it.
An assistant that answers from your documentation, always cites the source and says "I do not know" when it does not, with the accuracy rate measured before anyone is shown it.
See the situationAgentic operation guardrailsHow far can an agent go on its own, and who stops it?
The agent no longer just answers: it executes. Nobody defined how far it can go on its own.
What your agent can do on its own, what requires approval, what stops it and who answers when it acts wrongly. Built and tested with a real stop.
See the situationProduction model monitoringIs your model still working as well as it did on day one?
The model still answers. Nobody can prove it answers as well.
The measurement, the threshold and the shutdown criterion installed and handed over running to your team. We do not stay watching the dashboard.
See the situationAgent continuity planningHave you ever tested the failure plan for your AI agents?
The process already depends on the agent and nobody remembers how it was done by hand.
Plan B, written and actually tested: the agent gets switched off and how long your team takes to pick it up gets measured.
See the situationNon-human identity governanceWho answers for what your service accounts and your agents do?
We have service accounts with privileges and nobody knows who answers for them.
The inventory of everything that is not a person and has access, with a named owner, documented rights and its rotation and revocation rule.
See the situationThe other disciplines
AI strategy
12AI governance and compliance
9Data and analytics
5Cybersecurity and resilience
4Critical operations and crisis
5Mergers, separation and diligence
4Cost of AI and software
5Infrastructure and platform
5Or see it by the cycle: Vanguard, Forge and Taskforce. Or by the question you arrive with, in the seven questions. And if you already know the name, there is the full list.
What is the decision you have in front of you this week?