The terms we use about AI value, operating models, agents and governance, in plain words.
AI that carries out multi-step work on its own: it breaks down a goal, calls tools and systems, checks its output and retries, with a person specifying the task and reviewing the exceptions.
The work of making an organisation able to get lasting value from AI: deciding where it will pay, rewiring the workflows, roles and controls around it, and measuring the result in the P&L.
The controls, ownership and oversight that decide how AI is used: what is in use, what could go wrong, which controls apply to which risk, and which executive is accountable.
How prepared an organisation is to capture value from AI, assessed across data, technology, skills, governance and culture. Measured at the start of an engagement as a baseline.
Visible AI activity (chatbot rollouts, experiments, announcements) without the process, data and ownership changes that would make it pay.
Pushing data quality, integration and governance into a "phase two" that never arrives. Every shortcut taken at the start has to be repaid later, and the pilots built on top cannot scale.
How quickly a critical skill or a model advantage loses its value. Short half-lives mean training and model choices need a refresh plan, not a one-off budget.
The sequence in which AI takes on work: first assisting a person who stays in control, then doing routine work under review, then running defined work on its own with people handling exceptions.
The all-in cost of one completed unit of work, including compute, escalation to people and rework. The unit cost that still means something when agents do the work.
The share of work an AI system completes correctly first time, without a person correcting it.
A design where a person reviews, approves or corrects an AI system's output at defined points before it takes effect.
Sending each task to the model that does it well enough at the lowest cost. A routing decision that engineers make now sets a meaningful part of gross margin.
How an organisation turns effort into value: who decides, who does the work, with what tools, under what controls, measured how.
Optimising AI work for "we shipped something" rather than "we changed how the work runs". The portfolio fills with demos that never leave proof of concept. A form of AI theatre.
How many AI agents one person can productively supervise. A measure of how far work has been redesigned around agents.
A profit and loss account restructured by AI doing the work: human delivery costs fall, compute becomes a cost of goods, and new lines appear for model depreciation, AI liability and compute buffers.
The share of AI output that has to be corrected or redone. Often the hidden cost that breaks an agent business case.
AI tools used inside an organisation without its knowledge or approval. Finding it is the first step of AI governance.
Treating AI as a product to procure rather than a capability to build: buying the tool before the problem is defined, and choosing use cases because a vendor demoed well.
Estimating the value an AI use case can create, in pounds, and ranking it against the effort and risk of delivering it, so the first wave is chosen on value.
Rebuilding a workflow so people, agents, automations and systems of record pass work between them cleanly, with shared context, instrumented end to end and with feedback loops so it improves as it runs. Tools make individuals faster; orchestration makes the business faster.
Lumo's three-part approach, one part for each of its areas of strategy, enablement and execution: define the value, enable the organisation and realise the value.
A small function that tracks every live AI use case against its baseline in the P&L, reports value to leadership and reprioritises the next wave on the evidence.
The output of defining the value: which AI use cases to pursue, in what order, worth how much, owned by whom, with the baseline each will be measured against.
A fixed-scope first engagement of four to six weeks that takes an organisation from "where could AI pay?" to a costed first wave a board can approve.