What we build

Five things. Not twenty. Each is a category we ship repeatedly, which is why we can quote them fixed.

The short version: five things, not twenty. Voice agents, workflow agents, document and intake agents, multi-agent systems, and the AgentOps retainer that keeps them all working. Each one is a category we ship repeatedly, which is why we can quote them fixed.

Last updated: 19 August 2026

The five categories

Voice agents

: What it is. An agent that answers and places calls, holds a natural conversation, and completes the transaction — booking, triaging, qualifying, collecting — writing the result into your system of record.
: Who it’s for. Any business where the phone is still the primary channel and missed calls cost money: trades, clinics, dealerships, hotels, storage, logistics.
: What you get. Sub-1.2 second response, live calendar and CRM integration, escalation to a human with full context, transcript of every call.
: What it costs to run. AI voice agent usage costs typically range from $0.05 to $0.15 per minute on mainstream platforms — for example, Retell AI at $0.07/min — with premium tiers reaching $0.40/min (Retell AI pricing, 2026; Orvera AI, 2026). We model the actual usage cost per workflow in the audit.

Workflow agents

: What it is. An agent that watches for a trigger — a form, an email, a status change, a threshold — reasons about what should happen, and executes across your systems. A copilot answers when asked; a workflow agent acts on its own and escalates only when confidence is low.
: Who it’s for. Operations teams doing repetitive multi-system work: dispatch, scheduling, chasing, reconciliation, compliance checks.
: What you get. An entire monitoring-and-response loop removed, not just minutes saved per query.

Document and intake agents

: What it is. An agent that reads unstructured input — PDFs, drawings, emails, forms, photos, voice — extracts what matters, validates it against your rules, and writes structured data into your systems.
: Who it’s for. Anyone whose work starts with somebody re-typing information that already exists somewhere else.
: What you get. Intake without transcription, and validation before a human ever sees it.

Multi-agent systems

: What it is. Several specialised agents coordinating on a process too broad for one — one researches, one drafts, one validates, one executes — with a supervising layer managing handoffs and confidence thresholds.
: Who it’s for. Businesses past their first successful agent, ready to automate a process rather than a step.
: What you get. End-to-end process coverage. This is the second or third engagement, not the first.

AgentOps retainer

: What it is. The retainer. Monitoring, evaluation suites, model upgrades, integration repair, drift detection and workflow expansion.
: Who it’s for. Everyone with an agent in production. Non-optional in practice.
: What you get. Agents that keep working. Models get deprecated, APIs change, business rules move, and quality degrades quietly rather than loudly. Without evals running continuously, you find out from a customer.

AgentOps in detail →

Browse by job instead

These five are agent types. If you’d rather look at the work being done — predictive maintenance, after-hours calls, document intake, reconciliation, compliance evidence — see use cases.

Predictive maintenance →After-hours calls →Document intake →Compliance reporting →Reconciliation →See use cases →

What we don’t build

: Not a horizontal chatbot for your website — good products do that for less than we’d charge
: Not anything touching safety-rated control systems
: Not fully autonomous decisions in regulated processes requiring a human signature; we build to the approval gate and stop
: Not a project an off-the-shelf product serves better — we’ll name the product in the audit

Frequently asked questions

What’s the difference between a copilot and an agent?

A copilot responds when a person asks. An agent acts on a trigger and escalates only when uncertain. A copilot saves minutes per query; an agent removes a loop.

Can you build on a platform we already own?

Yes. Microsoft tenant, we build in Copilot Studio under your existing governance. AWS or Google, we build there.

Do we need multiple agents or one?

Start with one workflow. Multi-agent systems come later, once you have a working deployment and internal trust.

Who owns what you build?

You do. It runs in your accounts, on your data. End the engagement and keep it.