Robots and human relationships.

What follows says what the tools do here, what they are not permitted to do, why we took them up, and where they have failed us. It is written for the people who ask: volunteer directors, the lawyers who sit beside them, and administrators who have run these organizations for thirty years and want to know what has changed.

The line that does not move.

A person is responsible for truth, interpretation, dignity, professional boundaries, and release. Artificial intelligence is responsible for none of them.

No document leaves this company because a machine produced it. A named person has read it, judged it, and released it. Where something we have published is wrong, the failure is ours, and we will not offer the tool as an explanation.

What the tools may do, and may not.

They may gather the statutes, regulations and regulators' guidance that bear on a question, hold them in one place, and notice when one of them has moved. They may draft against research that a person has already settled, and they may check a draft against its sources. Every finding they touch is filed with the source it came from and the date it was read, and a register records what has been verified and what has not.

They may not settle a fact. Where two sources disagree, a person goes back to the governing document and decides. They may not release anything: a book, a page, a letter, or an answer by email. And nothing on this page is offered as a reason to trust an answer without its source, which is printed with it.

Who this is for.

Public Benefit exists for one person in particular. She agreed to serve on a board, without pay, expecting to help with fundraising and to look at a policy or two. What she found was staff to manage, payroll to answer for, and a body of legal and administrative risk that is genuinely difficult. Nobody described that to her when she said yes.

She is the scarcest resource in the Canadian charitable sector. Not money, and not staff. Leadership willing to serve for nothing, and to carry real financial exposure while it serves. Many directors meet legal and organizational complexity of this order for the first time after they have joined a board, and the invitation rarely says what they may be asked to carry.

The sector runs on those people. If they stop coming forward, it does not carry on with a shortage. It has to be rebuilt on some other footing, and nobody has designed that footing. Keeping her in her seat is therefore the most practical work available to us, and everything below is in service of it.

What we actually do.

Public Benefit is built on single source publishing. That is a technical writing method, and it is older than the current tools. One body of knowledge is settled once and kept in one place, and everything we publish is derived from it. A chapter in a book, a note at the back of that book, a page in a toolkit, and an answer to a question by email all trace to the same settled research. When a source moves, the research moves, and every piece of writing resting on it is flagged for checking again.

That knowledge base is built from primary sources: the statutes, the regulations, and the guidance published by the Canada Revenue Agency and the provincial regulators. Each finding is filed with the source it came from and the date it was read. Where two sources disagree, we go back to the governing document and settle it there, then keep testing what we settled.

We publish what is. We do not advocate for a position, and we leave it to the sector to decide what ought to change. That restraint is a design decision. A board deciding something hard is owed the ground it is standing on, and the decision on top of it is theirs.

Why we bother.

The rules grow denser every year. Federal and provincial bodies each regulate a part of a charity: incorporation, charitable registration, employment, privacy, health and safety, fundraising, tax. The consequences of getting them wrong are set at a corporate scale. A thrift store run by six volunteers carries obligations written with a warehouse in mind. People will tell you the regulators are likely to look the other way. Perhaps they are. It is a poor way to run anything, and an unfair thing to ask a volunteer to rely on.

The information that would settle most of this already exists. It sits across dozens of sites, written in whatever vocabulary each legislature happens to prefer, and it is expensive to keep current. Denominational bodies, membership associations and the larger charities each keep someone who tracks it. Mostly they work alone, and mostly they are doing the same work as the person one organization over.

Gathering dispersed material, holding it in one shape, and noticing when a piece has moved is what these tools are genuinely good at. That is reach, and it is not judgment. We use them for the first, so that a director can spend her evening on the decision in front of her rather than on finding out what the rules are.

Where it fails.

The most serious failure we have found is fabrication. Researching a question across provinces, the machine produced a Quebec statute that does not exist, because Ontario had one and symmetry looked more like an answer than asymmetry did. It was fluent, it was confident, and it was wrong. Nothing in the writing announced the problem. Only checking it against the source found it.

The smaller failures are constant. Citations that do not survive verification. Parallels drawn between things that are not parallel. A house style that took some twenty revisions before it stopped sounding like a machine, across three distinct voices for storytelling, technical writing and authorship. The first book was essentially written by hand.

So the checking is not a courtesy. When a source changes, a report names every piece of writing that now rests on stale ground, and that debt is paid before new work is taken up. A person reads the result and releases it, or it does not go out.

Set against that, the honest ledger. Work we would have expected to take about eighteen months took six, at the cost of building the checking apparatus and running it. A year ago none of this was possible. Six months ago it began to be.

Whose work this is.

The question we expect is whether this takes work from people. As far as we can tell, it does not.

The people who hold this knowledge are retiring. Charity administrators and network managers in their fifties, sixties and seventies are leaving, and what they know leaves with them. Everything we have began with one of them handing it over, and we have been fortunate in them, and in the people who have since given money, time and expertise.

Our work is to catch what is walking out the door, carry it forward, and put it where more than one organization can reach it. The result we want is not fewer people doing this work. It is more people asking better questions of each other.

How we treat people.

We write at a grade nine reading level, deliberately. Plain language is not a simplification of the law. It is the condition of the law being usable by the volunteer who has to act on it, and it is what makes translation possible, so that someone running a small charity here can read this in their first language rather than their third.

We settle the common ground first. The shared mechanics of running a charitable corporation are the same whether the organization is a congregation, a curling club, an arts council or a hospice foundation. We hold that base, and we leave the edge cases to the people who own them and know them better than we do.

Our books are priced low on purpose, and we answer questions. Some of the people who write to us are under real pressure, and a question answered this week is worth more to them than a book.

That is the line, and it is the first thing on this page because it is the one that does not move.

Paul B. Wolfe
President, Public Benefit Canada Inc.

Questions are welcome at info@publicbenefit.ca.