Project or retained
Software Development
Software your team can still maintain in a year.
Read the scopeProject or retained
Automation aimed at the work that actually costs you.
We apply Azure OpenAI, AI Search and Document Intelligence to specific processes. We start by telling you which ones are not worth automating.
Service
AI Consulting & Automation
Engagement
Project or retained
Assessment
2 – 3 weeks
The problem
Most AI projects solve the wrong process
The pressure to adopt AI usually arrives before anyone has measured which processes are expensive. What follows is a pilot that demonstrates the technology and changes nothing. The useful version starts from the other end: find the recurring, high-volume, rule-bound work, then ask whether a model is the right instrument for it.
Scope
What the engagement covers.
Anything outside this list is quoted separately rather than absorbed quietly.
Opportunity assessment: which processes are worth automating, ranked by hours and risk
Retrieval-augmented generation over your own documents, with Azure AI Search
Document processing and extraction with Azure Document Intelligence
Azure OpenAI integration into existing applications and workflows
Microsoft 365 Copilot readiness: data governance and permission hygiene first
Evaluation harnesses so quality is measured rather than demonstrated
Cost, data residency and tenant isolation review before anything ships
Deliverables
What you are left holding.
Every item here is an artifact you own, in your systems, readable by someone who was not in the room.
01
Opportunity assessment
A ranked list of candidate processes with estimated hours, risk and honest feasibility. Including the ones we recommend against.
02
Working pilot
Against your real data, in your tenant, with your permission model enforced.
03
Evaluation set
Test cases and scoring so you can tell whether a change to the prompt or model made things better or worse.
04
Production path
What it costs to run, who operates it, where it fails and what happens when it does.
Assessment: 2 – 3 weeks
Pilots run 4 – 8 weeks after the assessment, and only against processes that survived it.
Fit
Who this is for, and who it is not.
We would rather lose the engagement at this paragraph than three weeks in.
A good fit if
- You have high-volume document handling or repetitive knowledge work
- You are being asked to have an AI strategy and want one grounded in your own numbers
- You want Copilot deployed without exposing everything to everyone
Not a fit if
- You want a chatbot on the website
- You want a pilot that proves the technology rather than changes the process
Questions
About AI consulting and automation.
Does our data train the model?
Not with Azure OpenAI. Your prompts and outputs stay within your Azure tenant and are not used to train the underlying models. We confirm the data boundary in writing as part of every engagement, because it is the first question your legal team will ask.
Is Microsoft 365 Copilot worth the license cost?
It depends entirely on your permission hygiene. Copilot surfaces what a user already has access to, so in a tenant with loose SharePoint permissions it becomes a discovery tool for data people should never have seen. We assess that before recommending the spend.
How do you measure whether it works?
With an evaluation set built before the pilot, not after. If we cannot define what a correct output looks like, that is strong evidence the process is not ready to automate.
Often paired with
What this usually runs alongside.
These are the combinations that come up most often, not an upsell list.
Project or retained
Azure Infrastructure & Architecture
An estate you can rebuild from source.
Read the scopeRetained advisory
InfoSec Consulting
Security judgment, on retainer.
Read the scopeAI Consulting & Automation
Thirty minutes, no pitch.
Tell us what the environment looks like today. We will tell you what we would change first, and whether this is the right engagement for it.