We describe how your organisation actually works, then run software against that description — inside the systems you already have. Eleven universities and a vehicle assembly line operate on it today.
Businesses won’t win because they use more AI.
They’ll win because they become more intelligent.
Most organisations have spent two decades buying software. A CRM. An ERP. A helpdesk. A document store. A reporting layer, and the four spreadsheets that quietly hold the parts none of them cover.
Each system is competent on its own. None of them know about each other. So coordination falls to people — who spend the day carrying information between tools, re-entering what already exists somewhere else, and chasing the status of work instead of doing it.
The next generation of software will not replace these tools. It will coordinate them.
Every institution runs on a handful of things that matter and the rules about how they move. Most software buries those in features. We hold them as a description the software reads.
A university has applicants, programmes, documents, stages and a definition of “complete”. An assembly line has units, stations, checks and a definition of “passed”. Different words, identical shape.
So we write the words down as data, not as code. The description belongs to you and is readable by your own people. Change the description and the behaviour changes — no release, no rebuild, no ticket to us.
It is also why a second institution in your sector takes weeks. The engine underneath has never heard of an applicant, and it does not need to.
Higher education’s vocabulary, lit. The rest are other sectors’ — same engine, different description.
Decisions made, actions taken, answers returned, every one recorded.
Orchestration, grounding, guardrails, evaluation, channels, tenancy. Shared by every sector. It has never heard of a student — and a build fails if it ever does.
The objects your institution runs on, the states they move through, the rules that govern them, and who may see what. Written as data. Yours to read, yours to keep.
The ERP, the CRM, the document store, the four spreadsheets. Untouched.
Revenue execution that runs continuously: every enquiry qualified, enriched and routed the moment it arrives, follow-up drafted in your voice, and the pipeline kept current without anyone updating a record by hand.
Context-aware customer systems that answer from your policies, your history and your product — resolving the routine at any hour, and escalating cleanly, with the full thread, when a person is genuinely needed.
Processes that cross departments without crossing an inbox. Documents read, data extracted, approvals granted or held against your rules, exceptions surfaced early rather than discovered late.
The institutional knowledge held in contracts, tickets, wikis and people's heads, made retrievable in one governed, permissioned place — so that an answer takes a question rather than a search.
Operational data turned into a decision and then into an action. Forecasts that move a schedule, thresholds that open a ticket, anomalies that reach a named person while the number can still be changed.
The connective work underneath all of it — identity, permissions, audit, state and reconciliation across every platform you run — so the result is one system rather than a collection of scripts.
The engineering barely changes between sectors. The vocabulary, the systems of record and the regulator do.
The system identifies customers about to lapse and fires the right win-back offer automatically, before the relationship is gone.
Product copy, tags and search metadata generated across thousands of SKUs, in your brand's register rather than a generic one.
Tracking, returns and sizing handled around the clock, resolving the bulk of routine contact without a person in the loop.
Patient intake, insurance verification and scheduling automated, so front-desk staff stop working through a backlog of forms.
Conversations turned into structured, claim-ready notes — without the clinician charting late into the evening.
Patient questions answered and appointments booked at any hour, in your clinic's voice and within your policies.
Every enquiry qualified and routed the moment it lands, with the system scheduling on the spot rather than queueing a callback.
Listings, follow-ups and contract paperwork generated from a handful of details, ready for review and sending.
Personalised recommendations that put the right property in front of the right buyer instead of blasting the whole list.
Student enquiries qualified and routed to the right programme and counsellor within seconds of arriving.
Document collection, reminders and follow-up sequences run end to end, so nobody is holding the sequence in their head.
Assistants that extend the academic team's reach without adding headcount or fixed hours.
We built this — a university operating system, deployed across 11+ institutions.
KYC, contract review and report generation automated with a clear audit trail behind every decision the system makes.
Proposals and RFP responses drafted from your past work, your precedents and your tone — not a generic template.
Client questions answered from your policies and matter history, accurately, with the source attached.
Predictive scheduling and demand forecasting that flags an exception early enough for someone to do something about it.
Purchase orders, invoices and shipping paperwork processed without anyone re-keying a number that already exists.
A status agent that answers customers and partners instantly, at any hour, from the live state of the system.
We built this — a QR-code quality-control system for a vehicle assembly line.
Accounts researched, outreach personalised and meetings booked continuously, with a human taking over at the right moment.
Client calls turned into notes, tasks and updated records automatically, so context stops living in someone's memory.
We build the intelligence inside your own product, with our team as the engine behind it.
We built this — two commercial systems, end to end: one that runs sales outreach, one that turns account signals into a next action.
Rough numbers, honestly labelled. Move the sliders to your own reality.
That is roughly 4,860 hours a year — about 2.7 full-time people worth of capacity.
An estimate from your own inputs, not a promise. We put measured numbers against a real workflow on the call.
Artificial intelligence is changing software. But software alone does not change an organisation. Operational change happens when intelligence becomes part of ordinary work.
We are not, in the long run, an AI company. Intelligence is what this work takes right now; in a few years it will take something else, and we will build with that instead. The part that does not change is the part we are here for — making the thing run in production, on an ordinary Tuesday, with nobody watching a demonstration.
We work forward-deployed — from inside your operations rather than alongside them. Four stages, in this order.
We sit with the team and find the work that is expensive, repetitive and worth changing first. We also say which of it should be removed rather than automated.
One narrow system, wired into your real tools and your real data, built in your accounts on your infrastructure. No sandbox, no synthetic dataset.
It goes live with your team and your customers. We watch it under real load, correct what the plan got wrong, and harden it until it is boring.
We stay accountable for it: monitoring, iteration, and the next system once this one holds. The people who built it are the people who run it.
An operating system for university administration, deployed across more than eleven institutions.
A QR-code quality-control system built for a vehicle assembly line.
Two commercial systems built end to end — one that runs sales outreach, one that turns account signals into a next action.
Everything else is negotiable. These are not.
A pilot that never ships is a cost with a friendlier name. If we cannot see the path to production, we say so before you pay us anything.
You receive working software, running in your own accounts, documented, with your data exportable at any time — not a set of recommendations.
Your CRM, your spreadsheets, the ERP you are committed to: they stay. We build into the stack you have already paid for and already trained people on.
The people on the first call write the code. There is no account layer between you and the engineering, and nobody to point at when something breaks.
Some of it should be removed, not made faster. Saying which is which is part of the engagement, including on the days it makes the engagement smaller.
“Companies kept paying for AI strategy and receiving slides. The technology was ready. What was missing was someone who would build it, wire it into what they already ran, and stay accountable once it was live.”
Nithish Singh — Founder, Doing AI
Doing AI is a small engineering team by design. The people on your first call write the code and operate it afterwards. There is no advisory layer and no handover to a delivery organisation.
It is fair to ask what a small team means for continuity. Everything we build runs in your accounts, on your infrastructure, documented — not locked inside ours. Your systems and your data stay reachable regardless of who is available in a given week, and continuity is written into the agreement rather than left to trust.
Representative, not exhaustive — each engagement uses what its systems of record require. Model choice is made per task and stays portable, and nothing is built on infrastructure you cannot reach.
With one system, fixed in scope and fixed in price, against a workflow we have both agreed is worth changing. You approve the number before any work begins. Most organisations then move to a standing engagement in which we operate what we built and engineer the next system — which is where the compounding is.
Weeks. A first system is typically live in two to four, and in front of real users in four to six. The scope is deliberately narrow, so that the first thing you see is real rather than representative.
In your accounts, on your infrastructure, documented. Your data remains yours — exportable on request, governed, auditable, and never used to train anyone else's model. Commercial and continuity terms are set out in the agreement before work starts.
No. We build the system, integrate it with the tools you already run, and operate it. You do not need to hire for it or manage infrastructure to keep it running.
A consultancy is accountable for advice. We are accountable for software that runs. The engagement is an engineering partnership: we build inside your operations, stay on after launch, and are measured by whether the operation changed.
Thirty minutes. We will tell you what we would build first, roughly what it costs, and where this will not help.