Start a conversation

Our Approach

One model.
Four frameworks.

Consulting fails small organizations in a predictable way: the judgment walks out the door with the consultant. The Durable Operation is the model everything else hangs from — and these frameworks exist so the reasoning stays behind when we do.

Framework 01

The Durable Operation

The whole model on one page. Four layers of fundraising system, in the only order that works, held by two rails that apply from the first day rather than arriving later. Everything else we do is a component of this.

The Durable Operation — four layers of fundraising system, held by two rails Four stacked layers, bottom to top: The Data Foundation gives you a file you can trust; The Year on One Page gives you a year that holds; The Second Gift System gives you donors who come back; The Monthly Giving Program gives you revenue that renews itself. Two vertical rails run the full height: The Record, meaning everything is written down, and The Human Line, meaning the boundary between what AI does and what only a person does. Both apply at every layer. The Record WRITTEN DOWN AT EVERY LAYER The Human Line WHAT AI DOES — AND WHAT ONLY YOU DO LAYER FOUR Revenue that renews itself The Monthly Giving Program — roughly 3× the retention of a one-time gift LAYER THREE Donors who come back The Second Gift System — the ninety days that decide whether they give twice LAYER TWO A year that holds The Year on One Page — a calendar that survives the month everything goes wrong LAYER ONE — THE GROUND A file you can trust The Data Foundation — nothing above this works if the list underneath it is wrong Build from the ground up. The rails are not a stage you reach — they run through every layer, from the first day.
Why it's built this way: Small operations don't fail because the people are bad at fundraising. They fail because the layers were built out of order, nothing was written down, and the one person holding it left. The rails are the answer to the second and third problems.

Framework 02

The Human Line

Every task sits somewhere on two axes: how often it repeats, and how much judgment and relationship it requires. Plot the work honestly and the AI question stops being philosophical — it becomes an inventory you can finish in an afternoon.

The Human Line — a two-by-two matrix plotting work by repetition volume and judgment required High volume with low judgment should be automated. High volume with high judgment should be augmented — AI drafts, a human decides. Low volume with low judgment should be eliminated. Low volume with high judgment must be protected and kept entirely human. Protect Keep it entirely human. No drafts, no assist. THE MAJOR-GIFT ASK THE PRICING CONVERSATION THE APOLOGY THE STORY ONLY YOU CAN TELL BOARD & STRATEGY JUDGMENT Augment AI drafts. A named human decides and signs. SEGMENTED APPEAL VARIANTS PROSPECT & FUNDER RESEARCH BRIEFS GRANT DRAFTS AND BOILERPLATE PROPOSAL AND PITCH FIRST PASSES DONOR / CUSTOMER TRIAGE Eliminate Before you automate it, ask why it exists. THE REPORT NOBODY READS THE MEETING THAT DECIDES NOTHING DUPLICATE ENTRY BETWEEN TWO TOOLS THE EVENT THAT NETS ALMOST NOTHING Automate Build it once. Let it run. Check it monthly. GIFT RECEIPTING AND ROUTING FAILED-PAYMENT RECOVERY (DUNNING) LIST HYGIENE AND DEDUPLICATION SCHEDULING AND REMINDERS STANDING DASHBOARD REFRESH THE HUMAN LINE — EVERYTHING ABOVE NEEDS A PERSON Judgment & relationship HIGH LOW Volume & repetition LOW HIGH Rule of thumb: if you cannot name the human who signs it, it does not belong above the line.
Why it matters: Most AI conversations stall because they're framed as a values question. This reframes them as an inventory question — answerable quickly, and defensible to a board, a congregation, or a customer.

Framework 03

The 90-Day Second Gift System

Roughly three-quarters of new online donors never give again. The research is unusually clear about why: the first ninety days decide it, and most organizations have no deliberate sequence in that window at all. This is the highest-return thing a small shop can build, and it is entirely procedural.

The 90-Day Second Gift System — a five-stage donor welcome sequence Days zero to two: thank the donor like a human. Days three to fourteen: prove the gift landed, with no ask attached. Days fifteen to forty-five: invite rather than solicit. Days forty-six to seventy-five: make the second ask, framed as continuing. Days seventy-six to ninety: open the monthly giving door. DAY 0–2 Thank like a human A real note. Their name. Not a receipt. Not a form. DAY 3–14 Prove it landed One specific thing their money did. No ask attached. DAY 15–45 Invite, don't solicit A question, a story, a look inside. Earn the next ask. DAY 46–75 Make the second ask Framed as continuing, not starting over. Name gift one. DAY 76–90 Open the monthly door Convert best responders to recurring. Roughly 2× value. 19% first-time donor retention → 59% once they give twice The whole economics of your file turns on one gift. Acquisition costs $1.00–$1.50 per dollar raised. Renewal costs $0.20–$0.25. Retention is not a nice-to-have. It is the business model.
What you get: Five written templates, the trigger rules for who receives what, the threshold above which a human picks up the phone, and a one-page dashboard that tells you within a quarter whether it's working.

Framework 04

The Founder's Operating Loop

For founder-led businesses, the failure mode isn't refusing to adopt AI — it's adopting it enthusiastically in the one place that was doing the selling. This loop keeps the sequence honest: measure before you automate, isolate before you deploy, and document before you move on.

The Founder's Operating Loop — a four-stage quarterly cycle Instrument: know what actually drives revenue before automating anything. Isolate: inventory the work against the Human Line and pick one thing. Deploy: ship one small system with a human in the loop and measure it. Document: write it down so it outlives whoever built it. Repeat quarterly. 01 02 03 04 RUN THIS LOOP quarterly Instrument Know what actually drives revenue before you automate anything. Isolate Inventory the work against the Human Line. Pick exactly one. Deploy Ship one small system with a human in the loop. Measure it honestly. Document Write it down so it outlives whoever happened to build it.
The discipline: One loop, one system, one quarter. Businesses that try to deploy AI across five functions at once end up with five half-built things and no evidence about any of them.

What you actually get

Every position in the model has a playbook behind it.

The frameworks above are the argument. These are the working documents — the things that get filled in, handed over, and used after we've gone. Two of them are below in full.

Layer one

The File Audit

Your list scored on eight dimensions — duplicates, addressability, email coverage, lapsed identification — with the fix order. Ships with The Tech Stack Blueprint below.

Layer two

The Cadence Calendar

Twelve months of touches on one page, each with an owner and a date — plus the written rule for what gets cut first when a month disappears.

Layer three

The 90-Day Sequence

Five touches, the trigger for each, the copy in your voice, and the gift threshold above which a human picks up the phone instead.

Layer four

The Sustainer Ladder

Recruit, onboard, recover, upgrade — four rungs with the ask language for each, and the dunning rules that stop the file leaking quietly.

The rail

The Protect List

Five questions that decide whether a task may ever be AI-assisted. Any yes and it stays human. In full below.

The rail

The Binder Index

The list of everything that must be written down for the operation to survive a departure — logins, conventions, calendars, decision rules. Included in every system.

Playbook · The Human Line

The Protect List

The Human Line tells you the shape of the decision. This is how you actually make it, task by task, without a philosophical argument breaking out. Five questions, asked in order, with a rule at the end that anyone can apply.

The Protect List — five questions that decide whether a task may be AI-assisted Five yes-or-no questions. Does it go out under a specific person's name? Would the reader assume a human wrote it personally? Does getting it wrong damage a relationship rather than a metric? Does it need knowledge that only exists in someone's head? Is this the moment someone decides to give, stay or leave? Any yes means protect it and keep it fully human. All no means it can be augmented or automated depending on how often it repeats. ASK THESE FIVE, IN ORDER, ABOUT ANY TASK 1 Does it go out under a specific person's name? A signature is a promise that someone read it. 2 Would the reader assume a human wrote it personally? If they would be surprised to learn otherwise, that is your answer. 3 Does getting it wrong damage a relationship rather than a metric? A bad open rate recovers. A bad condolence note does not. 4 Does it need knowledge that only exists in someone's head? The detail that is not in any system is usually the detail that matters. 5 Is this the moment someone decides to give, stay, or leave? Decision moments are the whole relationship, compressed. ANY YES → Protect. A person writes it, start to finish. ALL NO → Augment or automate, by how often it repeats.
Use it like this: List every recurring task your team does. Run each one down these five questions in a room together. The disagreements collapse fast, because people stop arguing about AI in the abstract and start arguing about a specific email.

Playbook · The Data Foundation

The Tech Stack Blueprint

What to run at three budget levels, and the named condition that tells you it's time to move up. Most small organizations are either paying for a tier they don't need yet or holding on to a spreadsheet two years past its limit.

The Tech Stack Blueprint — what to run at three budget levels Three tiers. At zero dollars a month: a donation platform that does not take a cut, a shared spreadsheet as the donor list, free workspace and design tools, and a free email tier. At about fifty dollars a month: a real entry-level CRM, a proper email tool, and the same donation layer. At about a hundred and fifty a month: a CRM with retention analytics and automation, plus a matching-gift tool. Graduate a tier only when a stated trigger is hit. TIER ONE $0 / month Under ~300 donors Donation platform that takes no cut Donor list in a shared spreadsheet Free workspace + design tools Free-tier email, under 500 contacts GRADUATE WHEN The spreadsheet has duplicates nobody can confidently resolve. TIER TWO ~$50 / month ~300–2,500 donors Entry-tier CRM, unlimited users Real email tool with automation Same donation layer, now synced Receipts and tags handled for you GRADUATE WHEN You want to know retention by segment and can't work it out. TIER THREE ~$150 / month 2,500+ donors, staffed CRM with retention analytics Recurring + failed-payment recovery Matching-gift lookup at checkout Dashboards your board can read DON'T GRADUATE PAST THIS Until you have a staffed shop, enterprise tools are overhead. Categories, not brands — pricing and vendors change every year, and the right pick depends on what you already run. The rule that matters: never move up a tier on ambition. Move when a named thing has actually broken. And route giving through a channel you own, so the donor list is yours rather than a platform's.
Why no brand names: Vendor pricing and free tiers change every year, and the right pick depends on what you already run. The tiers hold; the logos don't. Specific recommendations come in the engagement, current as of that week.

Working principles

The rules underneath the diagrams.

AI drafts. Humans send.

Nothing carrying your name reaches a donor or a customer without a person reading it and choosing to send it.

Disclose with a reason, or not at all.

Disclosing AI use suppresses response — unless you attach a mission-connected reason, at which point the penalty largely disappears.

Narrow beats comprehensive.

A strategy naming five priorities has named none. Every engagement finishes one system.

The playbook is the deliverable.

Median fundraiser tenure is under two years. Anything living only in someone's head is a liability with a countdown on it.

Put it to work

Which layer are you actually standing on?

Twelve questions and about ten minutes will tell you — and you keep the readout whether or not we ever speak.

Take the free assessment See the six systems