The Hope Group · Lab Notes

Five questions to ask any AI vendor (we answer them too)

2026-09-21T00:00:00Z

Every organization we work with eventually has to sit across a table from a vendor and decide whether to trust a piece of AI. Most do not have a technologist in the room. You do not need one. You need five questions, and the willingness to sit in silence until they are answered.

These are The Hope Standard, our own framework for responsible AI, turned around to face outward. We answer them below, and we think every vendor should.

1. What do you collect, and what happens to it?

A good answer lists the specific data, why each piece is needed, where it is stored, who can see it, and when it is deleted. It fits on a page.

A bad answer starts with “industry standard” and ends with a link to a privacy policy.

Ours: We collect only what is needed, protect it rigorously, and never sell it. Before we build anything we write a data inventory, and if a field has no reason to exist, it is cut. Every product ships with a plain-language data page.

2. How do you know the system is making good decisions?

A good answer describes how decisions are logged, how often a person reviews a sample of them, and what was found last time.

A bad answer is “the model is very accurate.”

Ours: We test how our systems make decisions and document what we find, including the surprises. The findings go to the client in writing.

3. Who did you test it on, and what did you find?

A good answer names the groups the outcomes were compared across, the gaps that showed up, and what changed as a result.

A bad answer is “we did not find any bias.”

Ours: We examine our models and data for unfair outcomes before and after launch. If a gap appears, the fix ships before the feature does.

4. Will my staff and the public know when AI is involved?

A good answer shows you the label in the product and the sources the system used.

A bad answer is “it is seamless.”

Ours: Users know when AI is involved and how it reached its answer. Anything generated or ranked by AI is labeled and shows its sources.

5. Who decides?

A good answer points to a person, with a name, who can override the system, and explains what the system is not allowed to decide on its own.

A bad answer describes the workflow without a human in it.

Ours: AI assists. People decide. No final decision about a person’s money, housing, health, or job is automated, and a human’s name is on every recommendation that matters.

Print this. Bring it to your next vendor meeting. If the answers are vague, so is the product.

What "local AI" actually means for your organization's data

2026-09-22T00:00:00Z

When people hear “AI,” most picture a chat box that sends whatever you type to a company far away. For a lot of tasks that is fine. For a case file, a client list, a payroll question, or a grant budget, it is not.

“Local AI” means the model runs on a computer you control: a workstation in our lab, a server in your office, or the laptop already on your desk. Your data goes into the model and the answer comes back, and nothing leaves the building.

What it changes

Three things.

Where your data lives. With a cloud tool, every prompt is a small export of your organization’s information to a vendor. With a local model, the data never crosses the wall.

Who can train on it. Nobody. There is no vendor on the other end to build a product out of your client list.

What it costs. Electricity, not per-word fees. That sounds small until you try to automate something that runs a thousand times a day. Local models change what a small organization can afford to do.

What it does not change

Local models are smaller than the largest commercial ones. For some tasks, long reasoning, polished writing, unusual languages, a frontier model is still better. We use those too, on our terms: business agreements that prohibit training on your data, no retention beyond the request, data minimized before it is sent, and a label in the interface so you always know which model answered.

How we decide

The checklist we use in the lab fits on an index card.

Does the task involve information about a person? Local when we can, and minimized before it travels when we cannot.
Would you be comfortable pasting it into a public website? If not, local.
Does the task need the very best writing or reasoning? A frontier model, with the protections above, and a person reviewing the result.
Can the answer affect someone’s money, housing, health, or job? A person decides. The model only drafts.

What it looks like in practice

In our apps we are pioneering local models for the steps that touch sensitive data, starting where the task is small and the stakes are high. It is not yet our default for every task, and we say so: the largest commercial models still do most of the heavy lifting, under terms that prohibit training on your data. The steps that touch the public web, like Project Lookout scanning a procurement portal, can run anywhere, because that information is already public. The line is drawn around your data, not around convenience, and we move it outward as local models prove they can hold it.

If you want to see this running, come by the lab. We will show you the machine.

Why we stopped calling ourselves a consultancy

2026-09-23T00:00:00Z

For five years, if you asked what The Hope Group did, we would have said something like “AI consulting for the organizations that need it most.” That was true. It was also incomplete, and it described us by the shape of our invoices rather than the shape of our work.

Here is what we actually do. We build software. We test it against our own ethical standards before anyone uses it. We audit it for bias, and we keep auditing after launch. We publish how we build so other people can build the same way. Then, because we know what it takes, we help organizations adopt AI with the same care.

That is not a consultancy. That is a lab.

The difference matters

A consultancy sells hours. A lab makes things and learns in public. When a consultancy is done, it leaves. When a lab ships a product, it is accountable for that product for as long as people use it. We wanted the second kind of responsibility, so we are naming it.

It also changes what we say no to. A consultancy takes the engagement. A lab asks whether the work moves it toward the thing it exists to build. For us, that thing is a Digital Beloved Community: a world where technology expands human dignity instead of eroding it, where the people most often left out of the room help design the tools that shape their lives, and where no algorithm decides someone’s worth.

What stays the same

Everything that made people trust us. We are still the only City of Boston certified minority-owned AI business in Massachusetts. We are still the preferred AI trainer for the Black Economic Council of Massachusetts. We still start every relationship with a free conversation and no pitch, and we are still reachable months after the work is done. Training and technical assistance are not going anywhere. They are how the lab’s work reaches the organizations that need it.

What changes

We will publish more. The Hope Standard, our framework for responsible AI, is on this site in full, along with how we check each part of it. Every product will ship with a plain-language explanation of what it collects and what it does not. Our audits will be written for the people affected by them, not for a compliance binder.

We will build more in the open, and we will keep pioneering local approaches. A lot of the AI industry’s privacy problems come from a single habit: sending everything to someone else’s computer. We are exploring and offering the alternative inside our own apps, models that run on hardware we or our partners control for the steps where it matters most, because the organizations we serve carry data about people who have been let down by technology before. It is not our default for everything yet, and we will say so plainly as it grows.

Why “Digital Beloved Community”

Dr. King’s Beloved Community was never a metaphor. It was a description of a world where justice is real, poverty and exclusion are not accepted, and every person is treated as sacred. Artificial intelligence is now reshaping how people get hired, housed, served, and seen. It can deepen old divides or help close them. The difference is not the technology. It is who builds it, who it is built for, and what values guide it.

Responsible technology is not a slower path. It is the better one. We intend to prove it, one product at a time.