Engagements.
We help organisations put artificial intelligence into production: AI agents, chatbots and voice agents, retrieval-grounded systems built on their own data, and the data foundation underneath. Work is delivered by the principal directly. Not staffed out, not subcontracted.
We do the work we ship.
Plenty of data consultancies sell a methodology and staff it with whoever is free that month. That is not how this practice runs. Engagements are taken selectively, scoped honestly, and delivered by the person who scoped them: someone who has spent seventeen years inside enterprise systems of record and who now runs production software on the same principles.
It limits how much work we can take at once, and we are fine with that. If we cannot do a piece of work well, we say so at the proposal stage rather than discovering it at delivery.
Artificial intelligence, put into production.
We do not run AI pilots that never ship. Everything below is something we operate in our own products today, at our own cost, which is a materially different claim from having read about it.
AI agents & agentic workflows
Agents that take action rather than only answer questions: deciding when to follow up, sending the message, updating the record, escalating to a human when confidence is low. Includes agent design, tool and API wiring, scheduled and event-triggered runs, guardrails, and the audit trail that lets you see why an agent did what it did.
Typical trigger: a repetitive judgement call that a person makes fifty times a day, correctly but slowly.
Chatbot & voice agent integration
Conversational AI embedded into what you already run. Website chat widgets, voice agents on a real phone number, and messaging channels, connected to your booking system, CRM, or knowledge base so the conversation ends in an outcome rather than a transcript. Includes handover to a human and after-hours coverage.
Typical trigger: enquiries arriving faster than anyone can answer them, or arriving at 9 p.m.
Retrieval-grounded AI (RAG)
AI answering from your documents, policies, and records instead of from model memory, with a citation on every answer. Covers document ingestion, chunking and embedding strategy, vector and hybrid search, and the evaluation work that proves the answers are actually right before anyone external sees them.
Typical trigger: a chatbot that sounds confident and is wrong, or a knowledge base nobody can find anything in.
LLM integration & orchestration
Connecting language models to the systems that hold your data. Multi-provider routing with automatic failover so one vendor outage does not take you offline, model selection by task and cost, prompt and context management, structured output, and per-transaction cost metering with hard spending caps.
Typical trigger: an AI feature that works in a demo and falls over, or bills unpredictably, in production.
AI readiness & opportunity assessment
A written assessment of which problems in your business are genuinely tractable with current AI, which are not, what your data would need to look like first, and what each would cost to run per month. Delivered as a findings document, including a recommendation not to proceed where that is the honest answer.
Typical trigger: a mandate to “do something with AI” and no clear view of where to start.
AI evaluation & cost governance
Measuring whether an AI system is good enough to trust, and what it costs to keep running. Output quality evaluation against real cases, regression testing when models or prompts change, usage logging, and unit economics per conversation, per document, or per customer.
Typical trigger: nobody can say whether the AI is improving, or what it will cost at ten times the volume.
The foundation underneath it.
AI is only as good as the data beneath it, and most failed AI projects are data projects that were never done. This is the older half of the practice, and often where the real work turns out to be.
Data integration & migration
Designing, building, and stabilising the pipelines that move data between systems. Legacy platform migration, batch and near-real-time integration, reconciliation frameworks, and the recovery of pipelines that have become unreliable.
Typical trigger: a platform migration, an acquisition to integrate, or a nightly load that has started failing and nobody knows why.
Master data management
Establishing a single trusted version of customers, products, or vendors across systems that disagree. Match and merge rule design, survivorship logic, data stewardship workflow, and the governance model that keeps a golden record trustworthy after the project ends.
Typical trigger: the same customer existing four times across three systems, and no one able to say which record is correct.
Database administration & performance
Performance diagnosis and tuning, backup and recovery design, capacity planning, access control review, and upgrade planning for relational and distributed data stores. Available as project work or ongoing retained support.
Typical trigger: queries that used to take seconds taking minutes, or a recovery plan nobody has actually tested.
Data governance, discovery & privacy
Finding where sensitive data actually lives across an estate, classifying it, and evidencing how it is handled. Data mapping, retention and access review, and the controls that have to be in place before customer records are exposed to an AI system.
Typical trigger: a privacy obligation, an audit, or an AI project that has just discovered it does not know what data it is sitting on.
How we work together.
| Model | Suited to | Structure |
|---|---|---|
| Assessment | A defined question needing an answer before committing budget. | Fixed fee, fixed duration. Delivers a written findings document and a recommendation, including a recommendation not to proceed where that is the honest answer. |
| Project delivery | A defined outcome with a clear finish line. | Fixed scope and fixed fee against a written statement of work, invoiced on milestones. |
| Retained advisory | An in-house team that needs senior data judgement on call. | Monthly retainer for a set block of hours. Design review, architecture decisions, escalation support. |
| Operational support | A production data platform needing ongoing administration. | Ongoing monthly engagement with an agreed response commitment. |
On rates. Fees depend on scope, duration, and whether the work is project-based or retained, so we quote against a written scope rather than publishing a rate card. Every engagement begins with a no-charge scoping conversation, and every proposal states the fee, the deliverable, and the assumptions it depends on before any commitment is made.
What to expect.
1 · Scoping call
A conversation about the problem, the systems involved, and the constraint you are actually under. No charge, no obligation.
2 · Written proposal
Scope, deliverables, fee, timeline, and the assumptions the estimate rests on, all stated plainly in writing before anything is signed.
3 · Delivery
Regular written progress updates. Problems are raised when they are found, not summarised at the end.
4 · Handover
Documentation your team can operate from. We do not build dependencies on ourselves as a retention strategy.
Commitments we make
- Scope and fee agreed in writing before work begins.
- We will tell you when a piece of work is not worth doing, including when that costs us the engagement.
- Your data stays yours. It is not reused, retained beyond the engagement, or used to train models.
- Confidentiality agreements signed as a matter of course.
- Documentation delivered so your team can operate without us.
- Delivered by the principal directly. The person you scoped with is the person who does the work.
Sectors and geography.
Healthcare
Data platform work under strict privacy and retention obligations.
Retail
High-volume integration and master data across merchandising and customer systems.
Financial services
Formal grounding in capital markets and global market structure.
Remote and on-site
Engagements are delivered remotely as standard, with on-site work available where a project requires it. The company maintains a registered presence in both Maine and Washington State and has delivered work across Texas, Illinois, Oregon, and Maine.
Platform standing
BigID partner for data discovery and privacy governance; certified across Informatica and Oracle for enterprise integration, master data management, and database administration. Platforms we work in.
Tell us what is broken.
Describe the problem in a few sentences and we will tell you honestly whether it is something we can help with, and roughly what it would take.