Artificial intelligence
The pilot is the easy part
The AI inside an Australian organisation is usually a licence somebody bought. Governing it is not yet anybody's job. Ericom treats it as an operating problem: decide what it is allowed to touch, measure whether it is working, and widen it only once both of those have an answer.
Bought, not governed
The licences are already there. What is missing is everything around them.
Few organisations have decided which data the assistant may read, so it reads whatever the person asking can already reach, and that is usually more than anyone intended: sites shared for a project and never closed, folders opened to the whole organisation years ago, permissions that outlived the role they were granted for. What good looks like has rarely been defined, so there is no way to tell whether the tool is saving time or moving it to a different part of the week. And the tool has no owner, so after the first burst of enthusiasm usage settles into a handful of people using it well while most of the rest stop.
The technology did what it was asked to. What failed was running it as a service: a pilot shows that the model can do something, and a service settles all three before it is widened.
What we run
Scoped, governed and measured the same way whatever the AI is pointed at. The service is for Microsoft 365 Copilot, the AI you are already paying for.
Adoption is not usage
A licence count tells you what was bought. It says nothing about whether anybody’s week is different, and the two are measured in completely different ways. It is the first number that gets reported, and the second that tells you whether the programme was worth doing.
Common questions
Short answers first.
Do you build models?
No. Ericom's AI consulting works with the AI already embedded in the platforms you run, today Microsoft 365 Copilot, and governs how it is used. Organisations wanting bespoke models built and trained are better served elsewhere, and it is worth saying so before a scoping call.
Who keeps control of the tenant?
You do. Ericom has read access, reviews what it finds and recommends what should change; your team makes the changes under your own change control. Where you want the changes made for you as well, that is separately scoped project work.
How do you measure whether it is working?
Against what the work looked like before, agreed with you at the start and reviewed each cycle. That is why this runs as a service: a measure defined after the fact usually finds something to report, and the people asked to fund another year can tell.
What you already own
You are probably licensed for more AI than you are using. Tell us what is licensed, who is actually using it, and which of three questions you could not answer today: what it is allowed to read, whether it is saving anyone time, and who owns it. The first conversation is usually about governing and measuring that, not about buying anything further.
Talk to us about AI