Dear Fedes and P&I team —
Your team spends its days on land coming back to life — the science, the learning, the policy that make restoration real. And you count inspiration as a return before anyone counts money. We do relate to that…
What you're deciding is the question our whole space is wrestling with: what are people doing impact work supposed to do with these tools? Without giving up the judgement, the plural knowledge, and the long horizons the work runs on.
What does AI do to us? Currently, it doesn't look like it's making most of us more human, more loving, more joyful. Or that it is being used to foster coherence or positive impact, be it in terms of four returns or elegance.
Working out ways to do that in practice with a team inside the organisation aiming at a hundred million hectares would be a joy, though.
So here's our proposal — with a little more in it than you asked for: a prep call to hear what you're hoping for and worried about, then five working sessions — the first a hands-on opening workshop with Jo in Amsterdam — plus a frontier workshop with Krzysztof, and a landscape map with a journey sketched out that we'll hopefully tear up and redraw, together, from day one. Curious to hear your thoughts!
The short version. Over six months from September, your team builds its own way of working with AI — a landscape map we re-draw and score together, then a run of working sessions on your own material, deciding case by case what belongs. And what doesn't. What you keep is a practice you set yourselves, and a clear position on AI — including wherever that position turns out to be "nope".
Tools that support critical thinking rather than replace it, transparent enough to interrogate, that keep your analytical ownership, honest about their limits, with room to conclude some tasks are better done without AI.
How we work
We prepare, we refine together, you decide. We bring the research, the map, the tools and the examples; you take them apart on your own cases and keep what works, use what nourishes.
Your ToR already says what "better" means here: tools that support critical thinking rather than replace it, transparent enough to interrogate, that keep your analytical ownership, honest about their limits, with room to conclude some tasks are better done without AI. Everything we do is held to those five lines.
The landscape map you make with us
Your ToR's first ask is a landscape review. We make it a map, and we draw it with you.
It covers the tools and approaches that touch your actual work: research and literature mapping, synthesis and analysis, policy monitoring, reporting and communication, the impact measurement-adjacent tasks, and the internal, private, or owned workflows you might run yourselves.
We propose to literally map it out, visualize it, by intent. An overview to look at, with layers of depth and analysis to dive into — with white spots on the map that need exploring, and starting with the question how that map was made, where AI (which we will use) shows its hand, where it doesn't. To uncover what came from our brains, and what was done by silicon support crew.
A tool's place on it is set by what it's built for, who it really serves, what it costs to run, and whether it runs on fear or on something closer to trust. Nearly every big lab is built on the same thing — competition, who's biggest, who's fastest, who takes the market — so most of what you're handed is fear-based. We put that on the map in writing — plainly, so you can see where each tool's incentives really point. And we keep the map open: full of white spots, and questions none of us can answer yet. Maybe there's a promised land somewhere out there — a freer kind of AI, more private, more in your own hands, even something you could run locally and shape yourselves — and maybe there isn't. If we find it, we find it together, near the end, as one possible conclusion among a hundred. Nobody gets walked in a direction they didn't choose. The good open ones that already exist go on the map by name.
We score each tool against the things we think you care about: quality of sources; bias toward digital, English-language and Western knowledge; privacy and data exposure; cost and energy; what it does to your own judgement; how easily you can check its work; ownership and lock-in — And then we figure out together what YOUR scorecard looks like. That will be one of the instruments you keep — ready for every new tool that shows up after we're gone.
About a month before we start, the five of you and the five of us get on a call for ninety minutes — no slides, just questions. What you're hoping for, what worries you, what's off the table before we begin. And what would make this great? What would be the best possible outcome for each of you? That's what sets the compass for our map.
We come with the research done. Each of us — Jo, Carsten, Krzysztof — sends our own agents across the field, sets them to pick apart each other's findings, and we sort the pile by hand / brain / experience. You get the map already drawn and scored and our thinking on it — as a starting point to argue with. Then we redraw it together: a tool some of you already use goes up on the wall and we place it — where does it sit, what does it take from you. Someone runs a real literature search through it and we watch it pull ten English-language papers and miss the three that matter. That's the map being drawn — your work, your call, out loud.
The landscape, drawn with you
We come with a first pass drawn and scored — a starting point to argue with. What you're looking at here is a demonstration of the instrument; the real map gets drawn with your team. Click any tool to open its scorecard. And the dotted route? One example. It starts at session zero — the prep call — and we draw the rest with you: you set the direction, and change it at any point. Every session shifts to what you want.
Your scorecard, blank
We score each tool against the things we think you care about: quality of sources; bias toward digital, English-language and Western knowledge; privacy and data exposure; cost and energy; what it does to your own judgement; how easily you can check its work; ownership and lock-in — and then we figure out together what your scorecard looks like.
The sessions, on your real work
Each session runs two to two and a half hours on a live piece of your work: a short framing, a hands-on element, then an honest look at what the tool did to the thinking — what it saved, what it quietly cost, and whether it belongs in your practice at all. The team decides: keep it, change it, or drop it. Dropping it counts — your ToR asks for that result out loud. Those decisions stack up, session by session, into a charter you write: when AI helps, when you flag it, where a human signs off, what stays fully human.
The five sessions below are the route we'd propose through the map, one a month from September. We bring it as a draft — argue with it, reorder it, drop stops, add ones we missed. The syllabus isn't set in stone; we flow with where the team wants to go. If session three turns up something more urgent than session four's plan, session four changes. Between sessions there's homework — small things to try, so the learning keeps moving. Colleagues from other Commonland teams are welcome in the session that touches their work.
They run from the why to the what to the how: where you each stand on AI, what it does to real P&I tasks, and what — if anything — you'd want to own yourselves.
Opening workshop — the map, redrawn.
In person in Amsterdam, a full day or two half-days. Jo runs it. Coming in: the ninety-minute call behind us, real cases gathered, the first-pass map and scorecard drafted. In the room: how you use and avoid AI today, reflected back, skepticism treated as fair; the map argued over and re-sorted, tools moved, axes changed; the path chosen — breadth, depth, or owning something; the first charter lines written.
Your work, twice.
~90 minutes, online. One task, done the normal way — noting the judgement you used — then the same task with AI, watching what got faster, what got smoothed, where your own view dissolved into the model’s. This is where cognitive offload gets concrete: you see exactly what the tool takes off your plate, and what it quietly takes with it. Score both, reach a verdict, and re-claim what’s yours — keep what AI sharpened, drop what it flattened.
The literature map, without losing what matters.
~2–2.5 hours, online. A live research question. The baseline: what you found by hand. The AI pass: what it surfaced, what it invented or mis-cited, whose knowledge and which languages it skipped or down-ranked. The source-discipline line in your charter starts here, from what you watched happen.
Comparing documents without flattening them.
~2–2.5 hours, online. A real policy or donor document — the kind you’d drop into a tool to compare, summarise, or turn into a plain-language brief. We do it, and watch for what gets lost: context flattened, a local or Indigenous approach smoothed into a tidy Western summary. And we take the legal question head on — what the EU AI Act and GDPR actually mean when you paste a partner’s or a funder’s documents into a model, and which of these tools you can trust with them at all.
The finale — build or don’t.
Half a day, in person if the timing lines up. The verdicts recapped. One task from your own “what could be ours” list scoped end to end — what needs an agent, what’s a defined step, what stays human — and a working open-source version shown running on your material. You decide: build, pilot, or leave it. Then the charter, written live from every logged verdict, and the closing reflection. “Not worth building” is a strong outcome on its own.
Bonus — Krzysztof's frontier workshop.
On top of the five, and separate from them: Krzysztof comes in for one in-person session, open to colleagues from other teams. It's also the step-back session: how AI and agents change what a team like yours can actually do, and where the ground is likely to shift over the next six to twelve months — so your decisions aren't measured against last year's tools. Hard parts first — one real failure from your world, and the places you'd never want this — then the other side: research that turns up Global South sources you'd never otherwise reach, and local, open-source models running privately on machines you control, a hands-on lab on your own cleared material, honest about cost, limits and upkeep. We'll aim to run it when Krzysztof is in Amsterdam anyway, so it costs you nothing extra in travel.
Monitoring, reporting and your repeatable work run through sessions two to four as the cases you choose. This is the shape we'd start from; the opening workshop can rearrange it.
The charter, written as you go
The lines the team writes, one session at a time
These lines are illustrative — your charter fills with your own, session by session, from what you watch happen. A few we'd keep steady across everything:
- 1.Score each tool against your own criteria.
- 2.Reach a real decision every time.
- 3.Write your own view before AI touches the work.
- 4.Keep your sources straight.
And the lines you add — write them here.
Where it fails
Ask an AI research tool a real question and it returns a confident, ranked list — almost all English-language and already indexed. It reads like the whole field. Here, on an illustrative search, is what that leaves out.
What the tool found is real. What it could not reach stays invisible — the pull toward English and Western sources your ToR names, made concrete. In a session, this is exactly what we watch for, together.
The learning loop
If you're comfortable with it, we record the sessions. After each one the team spends half an hour naming what it learned, by hand. We fold the transcript, your reflection and the session material into a short learning note — and then you check the note. Did the AI catch it? Did it smooth over a disagreement? Did it sound more certain than the room felt? That check is part of the practice. The notes stack up into something Commonland can hand around later: a prompt log, a decision log, and learning notes in your own words.
You choose the path
Three shapes to start from — one, or a mix, decided in the workshop:
- breadth — one session per workstream, so the whole team moves together
- depth — take the workstream that hurts most and go deep, and leave a template the rest can copy
- owning something — less sampling, more building: find and prototype one private, open-source workflow you could run yourselves
And you set the working rules — by setting them, in the room, on real cases. What that looks like: in the research session the team decides "AI can draft a first literature list, but every source gets checked by a person, and anything non-English gets hunted down by hand." That line goes in the charter. It's yours, it came from watching the tool fail at exactly that, and it holds from then on. We keep maybe four lines steady across everything — score each tool against your own criteria, reach a real decision every time, write your own view before AI touches the work, keep your sources straight — and the rest is yours to set.
What is elegant.work?
elegant.work exists because I've been bootstrapping impact projects for a long time, and funding for bleeding-edge bioregional infrastructure barely exists. So we built a way to bring money into our space ourselves. The brilliant people in the impact field are mostly extracting from themselves — creating amazing impact projects they don't get paid for, the opposite of regenerative — so elegant.work is paid client work for exactly those people, and a bridge that carries regenerative practice into the mainstream. It's a Trojan horse, openly declared: whatever profit doesn't go into paying people or developing elegant.work goes into the elegant.earth and théra ecosystem — bioregional hubs, applied research, funding architecture for regeneration. A for-purpose for-profit, currently pointed at getting those hubs into the ground.
The name is load-bearing. Elegant, in the sense Aquinas gave it: integritas, consonantia, claritas — wholeness, harmony, radiance. Krzysztof and I talk about the same thing as companies and people being actually alive — self-actualizing instead of extracting from themselves. That's the standard the work is held to, AI included.
Commonland sits close to home. We're talking with other impact funds and organisations like Commonland right now, and they're all wrestling with the same question — what do we, as impact people, do with AI. We're hoping to make that a shared, ongoing thing for our space, and this programme would be a first brick of it.
Who you'd work with
Your ToR asks for one person, close to the work and hands-on. That's Jo. She leads this start to finish — the one in the room, running the sessions, holding the thread. Architect and systems thinker, twenty years in regenerative design, and the closest of us to Commonland's work — Listening to your Land is one of her best workshops.
The rest of us work behind her, and off the invoice. Jo designs the sessions and draws the first map with Krzysztof and Carsten — Carsten shaped the programme and stays close as it runs, sensing what wants to come out of it. Etrit brings the facilitation methodology. Alex builds the systems behind them, so what you try in a session actually runs.
So why a team at all, behind one facilitator? Two reasons. We want this — it's the work we most want to be doing, with a team asking the questions we think matter. And most of what the backend does is aligned with work we'd be doing anyway, so it isn't priced in: you're paying for Jo and the sessions, and the rest is there to give her the room to follow you wherever your AI discovery leads — deeper into research, into a build, into the legal questions, wherever your team wants to travel. We're glad to bring more when it helps, and none of it moves the price.
The five of us use AI hard and watch it closely. We've each concluded we need these tools to make the difference we want to make in the world — and we stay very critical of them, every single day we use them.
Jo has taught from real material for years — intensives, one-to-one work at hard decision points — and there's teaching depth behind her: Krzysztof runs his own courses and masterclasses, Carsten has been teaching since university, Etrit designs facilitation for a living.
One more thing about the five of us. We obviously choose to work with AI, a lot — and we question that choice every day. The energy cost and what these systems do to the world, yes. But just as much: what working with AI does to our own brains and our personalities. We'll dive into exactly that with you, and put our own answers up for challenge too. We're not going to sell you on AI. That's a promise.
This is close to the work we most want to be doing, with a team asking the questions we think matter.
What's yours at the end
- A landscape map of AI for your own work, drawn with you — and the scorecard you drew it with, ready for every new tool that shows up after we're gone
- Five tested sessions on real P&I material, each closed with a decision, plus Krzysztof's frontier workshop
- A short learning note after every session, checked by you
- A prompt and process log of everything that worked
- A decision log — use, adapt, refuse
- A practice charter, written by you and adopted by you: where AI belongs in your work, where it needs safeguards, where it stays out, and how you keep your own voice
- A closing reflection — what changed, what didn't, what's still open
- A grounded view on what, if anything, should travel wider in Commonland, with session-tested material for whoever it travels to
Schedule and fees
Six months from September — September matters, so the whole team can be in from the start. A call with the whole team, about a month out, comes first. One live session a month, five in all, with a small thing to try or chew on between, plus a bonus workshop from Krzysztof. The opening workshop in person in Amsterdam. Krzysztof's workshop in person when he's in Amsterdam. The finale in person if the timing lines up. The working sessions online. You're in Amsterdam, Jo in France, Krzysztof in Poland. Shared space in Notion, or wherever you already work.
One rate for everyone on this: €600 a day, the grid open. These are non-profit terms — the rate we hold for mission-led organisations. Part of why it's this low: with your permission, this becomes a reference case for us afterward.
Although this is a five-person team, the invoiced hours are mostly Jo's. The other team members step in when needed. Carsten will not be paid, given his relationship to the project lead for this assignment (Fedes). He is involved because this is the kind of work the five of us most want to be doing. Jo leads on her own merits.
Twelve days for a five-person team, six months, everything in — that works because we use AI intensely. The legwork runs on systems we built, so our days go into the thinking, the sessions, and the work only people do well. Getting there was the hard part. It's also what we're handing over — the parts that you might want from it, that is.
Afterwards, if useful: light support from ~€1,500 a month.
And a clean exit built in: after the landscape map and the first session you can stop — pay for what's delivered, keep everything. If it isn't working, that's a result too.
Ideas, if you ever want them
Your ToR is deliberately tight, and we've kept to it: a team building its own practice — not a strategy, not a content service, not a course. Everything above is that, and only that.
A few other things kept surfacing as we designed this, though — things we'd enjoy doing with you if they turn out useful. None of it is in the price, none of it is assumed, and none of it needs deciding now. We're laying them out because we think they could help, and because you should see the whole of what we're excited about. Say yes to any of it later, or never.
- A map that stays alive. The map you build is yours to keep, frozen. If you'd rather it kept current, our agents can keep refreshing it after we leave — each change a proposal you accept or wave off. Free through the first year, at cost after.
- The same map for other teams. If it works for you, any other Commonland team can make it their own — their parameters in, the agents keeping their map current. A later conversation, and only if it earns its place.
- A second map, for impact measurement. Zoomed into AI for the monitoring-and-evaluation side of your work — mapped and scored the same way as the main one.
- A short film, in your voice. Made with AI: a journey through the land of AI, the part that matters to you — in your voice.
- A course, from budget you already have. The core runs to €9,200 against your €12,000 ceiling; the roughly €2,800 left could turn what we build together into a self-paced course Commonland keeps — run across the organisation. Or maybe even handed to your partners. All that is a conversation for later, though
None of this changes the six months we're proposing. It's the fuller picture, laid out so you can steer.
How this answers your ToR
Phase 1 is the map we make together, scored against your own criteria. Phase 2 is the five sessions on your own work, with Krzysztof's workshop on top. The follow-up between sessions is the shared log and the small prompts. The learning notes and closing reflection are your decisions and the charter. And what your ToR rules out stays out: no IT or infrastructure work, no organisation-wide strategy, no AI writing your content for you.
What we'd ask of you
When the whole team can find ninety minutes for the first call. Which path you'd take, or whether you'd pick it live. Which work and which documents are safe to use in sessions. Who is in the core team, and which colleagues from other teams join which session. Whether sessions can be recorded for the learning notes. When Krzysztof is next in Amsterdam.
Your doubts, first
All five of you. One line each — the thing you'd most want to be wrong about, or the reason this whole subject makes you tired. They become the first material of the opening workshop. Skepticism is what this programme runs on; we'd like to start working with yours before anyone decides anything.
And we'd love to discuss and refine all of this in the follow-up conversation. Until then: leave your questions and comments directly in this document — it's built for that.
And one small thing about the page itself: if we do this together, this is where we'd do it. commonland.elegant.work becomes our shared workspace — the map lives here, the homework shows up here between sessions, you write on it and so do we. That's why it's a web page: somewhere we can actually work, together.
By the end it's yours — and "we don't use AI for that" is a fine place to land.