We optimise and streamline institutional processes.
We begin with your commercial objectives, then map your work products from the bottom up — across your data, workflows, dependencies and outputs — in a structured, logical and programmatic way. More often than not, the gain is in fewer steps rather than in new software, and in the overlaps between functions rather than inside any one of them.
The result is a decision-ready diagnostic and implementation roadmap, prioritised by commercial value, feasibility, data readiness, governance requirements and delivery complexity.
We begin with a paid Discovery. Tell us what you are trying to achieve commercially, and which functions are involved. We will return an initial assessment.
[email protected]Institutions accumulate processes and workflows that no longer resolve to any strategic intent. The work is to recover that intent, and to resolve the work to it.
That is what the name means. i²ntelligence — intent intelligence.
A testable operating model
Mapping turns a process into a model that can be tested: the inputs, decision points, controls, dependencies, outputs and verification steps that determine how the work is produced.
Timesheets and operational data are usually the best available basis for this, because they connect tasks to processes and to the measures the business already reports against.
Where value is created, lost or constrained
From the model we identify which work is chargeable and which is not; where bottlenecks form; where quality is constrained by the availability of experienced people; and which tools have already been bought but are not fully used.
Each finding is expressed in commercial terms, because that is the form in which it will be argued internally.
Where the larger unlock sits
A review of one process finds what is wrong with that process. A review of the whole finds the overlaps: the same data, the same controls, the same decision points recurring across functions, where one piece of work resolves several objectives at no additional cost.
These crossovers are usually where the largest gains are, and they are invisible from inside any single process. It is the reason the assessment is holistic and hierarchical rather than a pilot.
Every step is assessed
Each step is assessed against your strategy and deliverables: what can run without a person, what can be prepared in advance so that a person decides faster, and what must remain a judgement.
- RetireThe step comes out, and nothing takes its place. Many survive only because of a system since replaced, or a person since departed.
- AutomateThe step runs without anyone present.
- AssistThe work is prepared in advance, so that a person decides faster and more consistently.
- PersonThe step is left to a person, because it cannot yet be specified precisely enough to be anything else — and we will say so.
Defining a use case precisely
- what it must support;
- what it must be gated from supporting;
- the quality threshold it must meet;
- the evidence and research it relies on;
- how often that evidence must be refreshed;
- where human review is required; and
- what assurance gates and audit trails are needed.
Architecture you can hold to account
The systems that deliver automation and intelligence must remain maintainable from an accountability, cost and interpretability perspective.
You should be able to inspect outputs, audit which sources were used, understand where human review was applied, and justify the cost of serving and maintaining a model against the commercial value it creates.
Your data need not leave your infrastructure
Several architectures are available, and the choice is yours.
- an open-weights model hosted on your own infrastructure, which you own outright;
- work conducted inside a ringfenced environment over SSH or VPN, with the data remaining where it is;
- commercial APIs under zero-retention, zero-training terms; or
- anonymisation that preserves the statistical properties of the data while removing identifiers, including a bridge approach in which no client data reaches a model provider at all.
We will set out the trade-offs and build within whichever architecture your security and governance requirements permit.
How engagements are structured
Discovery
The application of the firm's method to a client's processes: a paid, data-driven diagnostic producing the roadmap.
Implementation
Building what the roadmap prioritises, using the firm's methods and systems.
Managed service
Operating and maintaining what has been built, where that is the sensible arrangement.
What we produce for you is yours: the diagnostic, the specification and the reasoning behind it. The firm retains the intellectual property in its own methods and systems. A managed service is a convenience, not a dependency.
What it involves
Discovery is the application of the firm's method to a client's processes: a paid, data-driven diagnostic across the functions in scope. It typically runs for three to six months, and closer to three where strong operational data already exists.
It requires access to timesheets and operational data across projects over time, at sufficient resolution to understand the tasks, the processes and their context.
Commercial arrangements, including risk-and-reward structures, are discussed before it begins.
For clarity
Discovery is a specialist, data-driven diagnostic and advisory engagement. It does not constitute production implementation, an audit, a formal assurance engagement, regulated professional advice, or the outsourcing, delegation or transfer to i²ntelligence of any operational, regulated, fiduciary, accounting, tax, legal, employment or management function.
What to expect
How much of a process will yield to specification is not known at the outset. That is why the work is conducted as research and reported as such, and why part of what is returned is the list of steps that could not be specified.
Where a fixed scope and a guaranteed answer are needed at signature, we say so at the outset and help you find the right form of engagement.
About
i²ntelligence is an AI research and development firm in London. Research is led by Amir Sani, PhD in machine learning and decision-making under uncertainty. Its research programme develops methods and systems for governed machine intelligence: deterministic, source-bound, auditable answering and routing; institutional process modelling from operational data; and decision-making under uncertainty. The firm owns the intellectual property in its methods and systems.
i²ntelligence builds and operates the instruments behind the Owner's Audit (intent map, demand graph, external assurance, reputation intelligence, commercial assurance) for owner-side clients contracted through AdapData.
The method has been applied across fund services and private markets, healthcare, hospitality, and cultural institutions. A portfolio of engagements in one sector does not make a firm a specialist in it. The method is the specialism.
Client work is confidential by default. We do not name clients, publish case studies, or quote performance figures.
Three outputs of the research
Three systems carry the programme into use. They share one discipline. No query-time model. Every answer cites its source, or it abstains. Deterministic and auditable.
- HotelCiteIntent resolution and governed routing over an evidence graph, with no model called at query time. Whether a compiled lexicon could reach the recall of a query-time model on unseen phrasings, without producing false routes, was not readily deducible. hotelcite.com
- Managed IntentOffline intelligence compiled to a deterministic runtime, with specified failure behaviour and demand learning held outside the answering path. No established method existed for learning from demand without letting that weighting distort relevance. managedintent.com
- OpenDataRoomPage-anchored document intelligence: verbatim answers tied to a page and a bounding box, with out-of-scope blocking and no generation at query time. Whether abstention could be decided without a generative judge was not known when the work began. opendataroom.com
Each is an output of the firm's own research rather than a finished commodity, and each is described honestly by status on the research page.
The method, written out
Eleven guides, each setting out one part of the method in the terms in which it is usually asked about.
- Business process mapping
- Process optimisation
- Automation consultants: what to expect
- Process improvement consultancy
- On-premise AI and self-hosted LLMs
- Human in the loop AI
- AI assurance and AI governance
- AI implementation roadmap
- Professional services automation
- Decision support and decision intelligence
- Owner-side instruments: how an asset is read from outside
Request a Discovery
Tell us what you are trying to achieve commercially, and which functions are involved. We will return an initial assessment.