Will AI make decisions for our team?

No. What we build reads, extracts, scores and drafts; a person on your team accepts or rejects every output, and the record shows who did. That’s how CF Labs applies the responsibility principle of the Montréal Declaration for Responsible AI: automation never absorbs accountability.

Is our data used to train AI models?

Not without your consent. Client data is scoped to the people who need it, never used to train models without consent, and deleted on request. Those commitments are published on our CF Labs page, so you can hold us to them before, during and after a build.

How do we know the AI got it right?

Test it before anyone relies on it. A pilot runs on your real documents, including the awkward ones: a superseded file, a missing year, two sources that disagree. Outputs show their source so a reviewer can check them, and the pilot passes only against criteria you set before the demo.

Where should a sustainability team start with AI?

With one task that eats hours every week and has a clear reviewer: reading supplier responses, pulling projects out of plans or drafting a recurring update. If you can’t name the task yet, the Data & AI Readiness Review ($6,500) helps you find it, along with the data it needs.

What is AI search visibility?

It’s whether AI assistants mention and cite you when buyers ask for recommendations. More of your buyers now start with ChatGPT or Perplexity, and the answer arrives somewhere you never see. Trovance, our own product, shows what the assistants say about you, why you’re missing and what to publish next.

Can Trovance guarantee we show up in ChatGPT?

No one can guarantee what an AI assistant says. Trovance shows you the answers every week, diagnoses why you’re left out, drafts the fix for a person to approve, then asks the question again to check whether it worked. Which engines it covers depends on your plan.

What does running AI cost the planet?

Compute has a footprint, and CF Labs tracks it the same way it tracks cost. It’s the discipline Council Fire asks of a client’s supply chain and has held itself to as a certified B Corp since 2010. Scoping the smallest system that does the job keeps both numbers down.

Can we switch it off?

Yes. Every system we ship keeps a way to switch it off, and the records underneath stay where they were, so your team can finish the work by hand. Prudence is one of the principles we build to: test on real data before anything runs unattended, and keep the off switch.

Practical AI for Sustainability Teams

CF Labs

AI that does the reading, so your team can do the deciding.

Prototype one AI task

Estimate your build

The problem

Your team spends its week reading, copying and reformatting.

Supplier questionnaires, hazard mitigation plans, grant reports, disclosure frameworks: sustainability work is buried in documents, and lean teams spend their days pulling numbers out of PDFs instead of acting on them.

A general chatbot doesn’t know which file is the approved one, and it shouldn’t be the last word on a claim you’ll publish. Meanwhile your buyers are asking AI assistants for a shortlist, and the answer arrives somewhere you never see.

What we build

6 kinds of build

Narrow, sourced and supervised

#1
Document review
Read a stack of plans, policies or supplier responses against your criteria and flag what’s missing, with the page behind every finding.
#2
Data extraction
Pull projects, figures and dates out of PDFs and spreadsheets into clean records, queued for a person to confirm before anything is imported.
#3
Internal assistants
An assistant built to answer from your approved sources and show which document it used, so staff stop hunting through shared drives.
#4
Drafting with review
First drafts of updates, briefs and grant narratives in your voice, with claims flagged for a reviewer before anything goes out.
#5
Scoring against your methodology
AI that scores material against your own framework. It’s the approach behind our Stakeholder Engagement Manager, now in development.
#6
AI search visibility, with Trovance
See what ChatGPT, Perplexity, Gemini and Claude tell buyers about you, learn why you’re left out and ship the fix. Trovance is our own product.

Introducing Trovance

Trovance · AI search visibility

A river through a misty forest, seen from above

A person at the gate

It reads, extracts and drafts. A person on your team approves what ships.

How it works

4 steps, in order

One task, tested, then trusted

#1
Pick one task
One document type, one reviewer, one output. A narrow pilot is how AI earns a place in the workflow.
#2
Test it on your messiest files
Real documents, including the superseded version and the record with a missing year, so you see how it fails before you rely on it.
#3
Keep a person at the gate
Every output waits for a named reviewer, and the record shows who approved what.
#4
Hand it over with an off switch
Your team runs it with documentation and training, and can pause it without losing the records underneath.