phase 02 · services
2 engagements open · q3 2026
Midnight Labs
home / services
services

Three areas of work. One way of starting.

We help organisations get better at learning from real work, not from another training catalogue. Most projects sit in one of three areas: how the ecosystem is designed, how teams use AI together, and how individuals exercise judgement when they use AI. We usually begin with a short, fixed discovery so everyone can see real signal before a larger build.

01 · ecosystem 02 · teams & AI 03 · judgement how we work
01 / ecosystem design

The three layers, designed as one system.

Every role sits inside an environment (physical and digital), a social world (who gets heard, who models what), and a technical layer (tools, data, workflows). We map those layers, then redesign the parts that are teaching the wrong habits. Where it helps, we connect your institutional knowledge to the AI tools people already use, using the Model Context Protocol (MCP).

typical shapediscovery → build
MCPwhen your stack fits
teamprincipal-led
your timescoped per phase
outputlive rituals + tools

What we do

We spend time inside one team or role. We look at how work actually flows: meetings, documents, handoffs, and the software in between. Then we change the smallest set of things that will shift behaviour, not a slide deck of recommendations.

MCP is optional plumbing: a standard way for AI assistants to read trusted internal sources at the right moment. We implement it when your organisation is ready to treat AI as part of the ecosystem, not a side experiment.

For sponsors outside engineering: it is how assistants stay aligned with what your organisation actually says—policies, approved sources, systems of record—so people get help that matches your standards instead of a model inventing your position.

What is different when we leave

  • → A shared picture of where your environment, social dynamics, and tools are teaching the wrong lesson—so you stop paying for training that the system quietly undoes
  • → A smaller set of owned rituals: how time is protected, how review actually happens, how handoffs land—so good behaviour is the default, not a heroic exception
  • → When it fits your stack, MCP so assistants draw on what you trust—policies, systems of record, approved context—instead of inventing your institutional voice
  • → A straight recommendation and next-step read: what to build, what to stop, and what would be wasted effort—even if you do not continue with us
02 · AI capability in teams

Shared habits for how the work uses AI.

Teams stall when everyone uses AI differently in private and nobody compares notes. We help groups agree on simple rules: when to use a model, how to check outputs, how to hand work between people and tools, and how to keep shared context so the room still thinks together.

typical shapeworkshops + practice
focusnorms + artefacts
teamprincipal + designer
your timeregular team hours
outputteam playbook + cadence

What we do

We run structured sessions with the people who share a workflow. We turn vague worry about "AI risk" into a small set of behaviours everyone can repeat: how drafts get reviewed, how sources are cited, what "good enough to ship" means when a model helped write it.

This often overlaps with ecosystem design. We list it separately because some clients only need the team layer addressed first.

What is different when we leave

  • → Explicit rules the whole workflow shares: when models are in play, when they are not, and what “good enough to ship” means—so nobody is guessing in silence
  • → A visible surface for AI-assisted work—drafts, checks, sources—so review is possible and blame does not pool in the room after the fact
  • → A steady rhythm with real examples (often weekly): the team adjusts the norms together instead of re-litigating the same mishap every month
03 · judgement & individual AI use

Building judgement when the model is always on offer.

Tools are fast; judgement is slow on purpose. We coach leaders and high-stakes roles on how to use AI without outsourcing thinking: asking better questions, checking outputs against reality, naming trade-offs aloud, and knowing when to ignore the shortcut. For a small cohort of senior leaders we also run Consilium, a private, year-long writing practice that sharpens point of view over time.

typical shape1:1 + cohort touches
Consiliumby application
teamprincipal
your timeshort regular blocks
outputpersonal practice + cues

What we do

We work with people whose mistakes are expensive: strategy, risk, legal-adjacent roles, senior operations. The work is not "prompt training" alone. It is practice in the memos, reviews, and calls they already live in, with feedback tied to how their judgement reads to others.

Consilium is separate from a consulting project: a quiet instrument with its own membership. Some organisations include seats as part of a longer partnership with us; others buy seats on their own.

What is different when we leave

  • → Personal defaults you can defend under pressure: when to lean on a model, when to slow down, when another human has to be in the loop
  • → Practice in the work you already own—memos, reviews, high-stakes threads—so judgement tightens where it actually shows up, not in a training sandbox
  • → Optional Consilium seat: a private writing discipline that keeps your stance legible to you over quarters, with peer reads that sharpen how your point of view lands—not a leaderboard
how we work

A steady rhythm, not a black box.

Most work opens with a six-week discovery: we learn your context, show early signal from what is actually happening in the job, and agree what to build. After that, deeper redesign runs in phases with a named owner on your side. If you want a lighter long-term relationship, we offer an annual partnership: a few structured check-ins a year, narrow help when things drift, and two Consilium seats for leaders when that fits.

step one

Six-week discovery.

Fixed scope and fee. You get a clear read on the problem and a recommendation before a larger commitment.

step two

Phased build.

We ship changes with the team that will run them. We stay until ownership is real, not borrowed.

optional

Ongoing partnership.

A few check-ins a year, light on-call support, Consilium seats for two leaders when that matches your bundle.

engagement model

What we always do, and what we skip.

we always
  • ✓ Work with a named internal owner from day one. No owner, no engagement.
  • ✓ Write so a tired reader can follow it in one sitting.
  • ✓ Tie claims to behaviour we can point to in how the job runs, not only in surveys.
  • ✓ Aim for visible movement inside the first six weeks of any live change.
we don't
  • ✕ Sell generic training programs or leadership theatre.
  • ✕ Run more than two deep builds at once.
  • ✕ Take work where no one will look at real outcomes or how the job actually runs.
  • ✕ Optimise for committee slides over one accountable sponsor.
pricing

Discovery is a fixed fee. Deeper work is quoted after we have shared facts on the table. Typical redesign work sits in the mid six figures AUD in Australia; ongoing partnership is an annual retainer. If budget is the main question, send a brief and we will say plainly whether there is a fit.

book sparring · APAC → book sparring · Europe submit a brief

Prefer email: tom@midnightlabs.com.au · janna@midnightlabs.com.au