01 · AI consultancy

AI that earns back hours.

We map AI to the part of your business that actually moves the number, then we build the thing that runs. No theater, no demos that never ship.

What we build

Four kinds of AI work.

01

AI strategy & roadmapping

Where AI actually fits in your operation, and where it doesn't.

We spend the first week mapping your workflows, data flows, and where time and money leak. Then we recommend a sequenced plan, the smallest set of AI levers that move your real number first.

  • Operational audit & data inventory
  • Lever prioritization & sequencing
  • Tool selection & buy-vs-build calls
02

Workflow automation

n8n, custom, or whatever fits — boring on the surface, transformative underneath.

We connect the tools you already use, automate the repetitive parts, and instrument the result so you can see what was reclaimed. Built to be handed back to your team.

  • n8n flows + custom serverless
  • Tool-to-tool integrations
  • Operational dashboards

Typical: 10–20 hrs/week reclaimed

03

AI agents & chatbots

Trained on your actual operations, not a generic FAQ pretending to be one.

Customer-facing or internal, agents that qualify leads, handle bookings, surface knowledge, or run process. We design the conversation, train on real data, and ship the deployment.

  • Lead qualification & booking agents
  • Internal knowledge copilots
  • After-hours customer chat

Typical: handles 60–80% of repetitive inquiries

04

System & data integrations

The unglamorous backbone that makes the rest of it actually work.

Connecting your CRM, billing, ops tools, and data warehouse so AI has real inputs and outputs that land in the right place. The work nobody brags about, which is why it pays.

  • CRM, billing, and ops connectors
  • Webhook & event-driven flows
  • Data warehouse pipelines

Productised use cases

The fast paths we ship often.

Two AI builds we've productised because we ship them often enough to price them flat. Bespoke work still welcome.

Honest calibration

When AI is the lever.
When it isn't.

Works well for:

  • Repetitive, rule-based workflows
  • High-volume customer questions with stable answers
  • Research and data-gathering loops
  • Integration between siloed tools
  • Patterns hidden in operational data

We'll tell you not to:

  • Process isn't documented yet, fix the process first
  • Volume is too low to justify the build
  • The "AI problem" is actually a hiring or training problem
  • You need it deployed this week

“We turn down AI projects we don't believe in. Saves both of us time.”

How AI engagements run

From map to running system.

01

Week 1

Operational audit

We sit with the workflows that hurt. Data, time-on-task, the fragile dependencies, whatever shows us the lever.

02

Weeks 2–3

System design + tool selection

We design the agent or flow on paper, pick the right stack for the job, and confirm scope. No demos that never ship.

03

Weeks 3–10

Build, integrate, instrument

Real build. Real integrations into your existing tools. Measurement wired in from day one so reclaimed time is visible.

04

Ongoing

Monitor, tune, expand

We watch the system in production, tune what drifts, and surface the next lever when the current one is steady.

Engagement shape

Right-sized, plainly priced.

Sprint

2–4 weeks

from$2,500

Best for: One focused automation or agent build.

Includes

  • Discovery
  • Build
  • Deployment
  • Training handoff

Partner

Ongoing

from$2,500/mo

Best for: Continuous expansion + maintenance + monitoring.

Includes

  • Monthly capacity
  • Ongoing tuning
  • Priority support

Approximate ranges. We confirm scope and number on the intro call.

What we build with

We pick the right stack for the job, not the trendiest one.

n8nOpenAIAnthropicMakeZapierLangChainSupabasePineconeVercel

Questions

Common AI questions.

  • Yes. Code, prompts, integrations, and configurations are yours, handed over at the end of the engagement. We optionally stay on to maintain, but ownership is never our leverage.
  • We architect around vendor risk, abstractions where it matters, monitored fallbacks where it doesn't. If a vendor changes terms, we swap the component without rebuilding the system.
  • Never used for training. Data stays in your environment when it can; when a third-party model needs it we use enterprise endpoints with no-training clauses. NDA up-front, audit log on every flow.
  • Built for it. Code is readable, integrations are documented, prompts are versioned, and we run a training session at handover. Most of our partners take the system in-house within 6 months.
  • Better. The best AI work plugs into your existing system of record rather than replacing it. We've integrated with HubSpot, Salesforce, Pipedrive, Airtable, Notion, Slack, Stripe, Shopify, and a dozen others.
Ready to talk

Ready to talk AI?

Bring the workflow. We'll tell you straight whether AI is the right lever, and if so, what it'd take to ship it.

Stay in the loop

Notes from the build, every two weeks.

Articles, opinions, and useful links on AI and the work.

No spam. Unsubscribe anytime. See past issues