Engagements · Research Stack Setup

We build the stack. Your team operates it.

An 8-week engagement where we deploy a working agentic research pipeline on your data, train your team, and hand the keys over. Not consulting. Not advisory. A built thing that runs.

Kickoff call within 24h of payment. Production cutover by end of week 8.

Request a scoping call See the 8-week timeline

Deliverables

What you get on day 56

Working agent pipeline

Running in your environment, on your data, owned by your team.

Validator rubric

Configurable pass/flag/fail criteria tailored to your standards — same shape as the validator behind our public research.

Runbook + ops doc

Version-controlled in your repo. Day-1 operability for anyone on your team, not just the people who watched us build it.

Incident playbook

The three most likely failure modes (rate limits, drift, validator regressions) with the exact remediation steps.

30-day post-handoff support

Included in the project fee. We sit in your Slack, answer questions, fix the things that break in the first month of real use.

Optional ongoing retainer

Quote separately if you want monthly optimisation, new agent additions, or scheduled validator-rubric reviews.

Who this is for

Three buyers we land hardest with

If you recognise yourself here, the 8-week engagement and $25k–$80k range were designed with your operating reality in mind.

In-house research team · 2–5 researchers

Team doing the same workflow repeatedly

Your researchers spend half their week on the same upstream collection + summarisation steps. They'd rather spend it on novel synthesis the agents can't do.

Outcome: We automate the repeatable layer. Your team keeps the strategy + judgment work; the grunt work runs while they sleep.

Funded startup · Series A/B

Research is a known need; hiring doesn't pencil out

You need research infrastructure but the first research hire is $180k+ all-in plus six months to ramp. The work is broadly known but not deep enough to justify a senior hire yet.

Outcome: Build the stack instead. ~$50k once, ~$0 ongoing inference cost. You hire the senior researcher in year 2 when the stack has proven the need.

Corporate research / strategy / intel function

Modernising from Word docs + Google searches

Your strategy or market-intel team currently does research by hand. The output is good but slow, and competitors have started shipping AI-augmented research at 10x your cadence.

Outcome: Agentic stack on top of your existing knowledge base. Same humans, same standards, ~5-10x the output cadence within a quarter.

Timeline

Week-by-week, from kickoff to keys

What each week looks like, in writing, before you sign anything. If we miss any phase by more than 2 business days without prior agreement, the refund window opens automatically — see services terms section 3.

1
Week 1 · Discovery
5 business days
Kickoff call + workflow diagnosis
- Founder-led kickoff call within 24h of payment
- We shadow your current workflow (1-2 hours of recorded screen-share)
- SOW finalised with scope + deliverables + 8-week timeline
- First draft of the architecture circulated by EOD Friday
2
Weeks 2–3 · Design
10 business days
Pipeline architecture + validator rubric
- Agent pipeline architecture confirmed against your data sources
- Validator rubric agreed — what passes, what gets flagged, what fails
- First end-to-end test run on a frozen sample of your data
- Weekly Friday call to walk through the test outputs
3
Weeks 4–6 · Build
15 business days
Agents deployed in your environment
- Agents deployed to your AWS / GCP / Azure (or ours with clear handover plan)
- Validator wired to the pipeline; first 50 real runs reviewed
- Daily standup (15-min) for the duration of the build
- Cost monitoring + drift detection wired in from day one
4
Week 7 · Train
5 business days
Your team hands-on + runbook delivered
- Hands-on training sessions for the people who'll operate the stack
- Runbook + ops doc written, version-controlled in your repo
- Incident playbook for the three most likely failure modes
- Stress test against last quarter's hardest research question
5
Week 8 · Handoff
5 business days
Keys handed over, production cutover, support window opens
- Production cutover during your nominated window
- All infra credentials transferred to your team
- 30-day post-handoff support window opens (covered in the project fee)
- Optional retainer quote for ongoing optimisation — separate scope

Pricing

$25k–$80k, scoped per project

Fixed quote in the SOW before any work starts. Milestone billing (50% kickoff / 50% delivery) available for any quote above $10k. Final number scoped against your data sources, the number of distinct agent pipelines required, and whether deployment is in your environment or ours.

~$25k — Single pipeline

One agent pipeline on one data source. Standard validator rubric. Your environment.

~$45k — Multi-source

2–3 agent pipelines across multiple data sources. Custom validator rubric. Standard cadence.

~$80k — Full research function

4+ pipelines, custom validator + auditing layer, hands-on training across multiple teams, post-handoff retainer ready to spin up.

You own everything

Code, prompts, validator config, runbook — full IP assignment on delivery, deployed in your environment.

7-day refund window

If the week-1 architecture draft doesn't fit, full refund within 7 days of that deliverable.

Daily standup during build

15-minute daily standup weeks 4–6. Shared Slack channel. Status email at every phase transition.

Full ten-section version (scope, IP, refund mechanics, confidentiality, payment, termination, governing law) at /services-terms →

Start with a scoping call

30 minutes, no obligation. Walk us through your current workflow and what you'd want operating differently in 8 weeks. We'll tell you within 2 business days whether it fits, what the rough quote looks like, and what the SOW would specify.

Request a scoping call