Emeris — AI capability that survives contact with operations

Key judgements

Where using AI and deploying it diverge

Five judgements about how organisational AI capability breaks down, each with a confidence level attached. These are the judgements I use to work out where capability is failing or building risk, set out in full with the six-pillar method behind them.

01Fluency with a chatbot is being mistaken for organisational capability
02Automation justified on reducing headcount caps its own value
03The people who have to live with the controls were never asked to design them
04Delivery went well and the capability left with the delivery team
05A restructure can break a capability that was working perfectly well

Read the five judgements

The work

Three situations Emeris is usually called into

If one of these describes your organisation, that is the conversation to have.

Your AI programme has stalled and nobody can say why

Diagnosis — work out what is actually wrong

A structured examination across the six pillars, ending in a diagnosis with a priority order rather than a list of everything that could be better. Often the answer is in the data — where it came from, whether it is fit for the decision being made with it, whether the lifecycle around it is disciplined enough that a model can be rebuilt rather than rediscovered. Just as often it is somewhere nobody has been looking, because the model was assumed to be the problem and the model was fine.

You have a pilot that works and cannot get it into production

Production — get one thing into production and keep it there

Taking a defined capability from prototype to something the organisation can run, govern and defend — including the parts that decide whether it survives contact with operations: escalation, monitoring, ownership, and who answers when it goes wrong. Documentation and knowledge transfer are deliverables, not a courtesy at the end. The engagement is designed to finish.

Your executives cannot tell real capability from a convincing demonstration

Judgement — build it in-house

Teaching the people who commission, build and sign off AI how to assess what they are being sold and what they already own — using methods drawn from intelligence practice rather than vendor material. The point is that you stop needing me.

In regulated financial services the same questions arrive through a prudential standard rather than a stalled programme. What CPS 230 now asks of an AI capability.

The practice

Where this comes from

Emeris is built on sixteen years in signals intelligence followed by a decade delivering production AI inside Australian government operations — systems that make decisions about real people, under real accountability, and are still running on Monday.

The less flattering half of the record is the more useful half. I have killed projects that should not have shipped. I have been in the room when a production model degraded and the escalation path turned out to be theoretical. Having owned a system while it was going wrong is the part that is difficult to fake, and it is the part that shapes how Emeris works.

The practice, and why independence matters

Start here

Thirty minutes to find out whether this is worth continuing

No deck, no discovery process, no proposal arriving three days later unless you ask for one. It is a diagnostic conversation, not a sales pitch.

You describe the problem — what has stalled, what the deadline is, and who is exposed if it does not land.
I ask about the parts people usually skip — where the data came from, who answers when it goes wrong, and what happened the last time it did.
You get an assessment on the call — what I think is actually wrong, based on what you have told me, whether or not you engage Emeris.
I tell you if it is not for me — including who would be better, if I know them. A wrong engagement costs you more than it earns me.

Tell me what isn't working Or email directly

Thirty minutes is usually the difference between a description and a diagnosis.

Line of Sight

The thinking, in writing, every fortnight

A fortnightly brief for people who have to answer for AI rather than announce it. One argument at a time, worked through properly — what is going wrong at the production boundary, why, and what it means for whoever is accountable. No news roundups, no vendor commentary, no predictions I would not put my name against.

Read Line of Sight Fortnightly · Unsubscribe any time · No list sharing