Transparency

How we use AI — out loud.

We build AI for a living. If we don't hold ourselves to a high bar on transparency, we can't expect anyone else to. These are our public commitments: how we label our own content, how we oversee the AI that touches our work, and how we stay in the conversation about what's right.

Version 1.4 · Updated February 5, 2026

Why this page exists

We can't ask clients to trust AI they can't see.

Every AI consultancy talks about responsible AI. Most keep the details private. We don't. If you're going to hand us work, you should know exactly what AI touches that work, who reviews it, and what happens when something goes wrong. This page is a living document — we update it as our own use of AI evolves, as regulations change, and as clients push us for more clarity. Version history lives at the bottom.

Commitment 01 · Labeling

Every piece of content we ship carries a label.

Emails, slide decks, deliverables, social posts, case studies — all of it. Four categories, applied honestly, with a human accountable at every level.

01

Human Generated, Human Approved

Content created entirely by humans, without AI assistance. Reserved for the highest-stakes work where the full weight of human judgment is warranted.

Used for: complex contracts and legal documents, sensitive client comms
02

Co-Created with AI, Human Approved

Human-AI collaboration with human direction throughout. A human writes or edits the bulk of the work; AI assists with drafts, research, or polish. A human takes full accountability.

Used for: most strategy docs, proposals, code generation, this website
03

AI Generated, Human Approved

AI produces the initial content; a human reviews, edits, and approves before publication. Human accountability is never removed — the reviewer is responsible for every word.

Used for: draft reports, first-pass summaries, outreach templates
04

AI Generated

Automated AI responses for low-risk, bounded scenarios — guided chatbots with specific data, scheduled updates, routine classifications. Built with guardrails, monitored, auditable.

Used for: FAQ bots, automated status pings, enrichment pipelines
Commitment 02 · Oversight

AI needs humans in the loop. So we keep them there.

Labeling tells you what happened. Oversight is how we make sure what happens is worth labeling.

01

Human Oversight

A human reviews every AI output before it leaves HQ, full stop. For automated AI, humans own the design, the monitoring, and the off-switch.

No exceptions. The human review is the floor, not the ceiling.
02

Quality Data

We curate training and context data deliberately. AI works only as well as the data underneath — so we treat data hygiene as a core quality gate, not an afterthought.

Data gate: a readiness review before anything hits production.
03

Rigorous Review

Regular monitoring of AI outputs for accuracy, tone, and bias. Issues are logged, tracked, and shared with the team — not buried.

Cadence: monthly internal reviews, quarterly cross-team audits.
04

Employee Training

Our team is trained on responsible AI use on day one and continuously after. We update training as the tools change — which is often.

Tracked: logged per person and refreshed every single quarter.
Commitment 03 · Engagement

Transparency isn't a policy document. It's a conversation.

We share our practices publicly, talk about them with clients, and update them when we learn better.

01

Proactive Sharing

Our AI practices live on a public page — this one. No clicking through four layers of legal text. If a client, peer, or reporter wants to know how we use AI, the answer is one URL away.

02

Industry Leadership

We share our commitments in public because we want other firms to do the same. Responsible AI isn't a competitive moat — it's table stakes. The sooner more firms show their work, the better the industry gets.

03

Open Dialogue

If you have questions, concerns, or a better way to do this — we want to hear it. Email us, call us, press us. Our public position won't change without a public update.

04

Regular Updates

AI moves fast. These commitments move with it. We publish a new version when our practices change materially — with a clear note on what changed and why.

Version history

What's changed.

A short log of material updates to these commitments.

v1.4 · February 5, 2026

Added examples of content that typically falls into each labeling category. Tightened language around oversight and review cadence.

v1.3 · October 12, 2025

Added a named contact for transparency questions. Moved the commitments page to the main site footer.

v1.2 · June 18, 2025

Introduced the "AI Generated" category for low-risk automated scenarios with explicit guardrails.

v1.1 · February 3, 2025

Clarified that human accountability is never removed, even when AI generates the initial content.

v1.0 · September 2024

Original commitments published.

Think we can do this better?

We mean it — transparency is a conversation, not a broadcast. If you know a better way to do this, or think we're getting something wrong, tell us.