I founded a second company. AutoGrant is a business distinct from Centered Networks even though the two are tightly coupled, built for the same organizations, and grounded in the same conviction about where AI is going. You can see it at autogrant.ai.
Two decades of experience sit behind this. I co-founded Collaborative Standards and helped launch AngelPoints, both of which were acquired by Blackbaud, and I have led Centered Networks as a Microsoft Solutions Partner since 2004, serving organizations like The Asia Foundation and Children’s Hospital Los Angeles. Before that, I spent time as Director of IT at The San Francisco Foundation, which gave me an inside view of how grant decisions actually get made. The organizations doing the most important work in this sector have been underserved by software for a long time. Both of these companies are a bet on closing that gap.
Centered Networks and AutoGrant point at the same customers and the same mission, but they are two genuinely different businesses, with different products, different economics, and different ways of being bought. Each one is stronger as its own thing. And the reasoning behind that structure is a bet about the technology itself: the era of the single, general-purpose assistant is giving way to an era of vertical AI, where the most useful systems are teams of specialized agents built for one domain. AutoGrant is what that looks like for grant funding.
Why two companies instead of one.
They are not the same business. Building them as one would have made both worse.
Centered Networks is a Microsoft Solutions Partner. We govern the platform a mission-driven organization runs on: identity, devices, email, data protection, and the Microsoft 365 and Azure foundation that makes AI safe to turn on. We are bought by an IT lead, a CFO, or a board, and we are paid to operate that platform as a managed service over years. The unit of value is a governed environment and the accountability that comes with it.
AutoGrant is a software product with a service around it. It is bought by a development director, an executive director, or a grants team, and the unit of value is a funded application. Its pricing reflects that: the core model is a success fee charged only on awarded grants, with no subscription floor and no upfront cost. That is a fundamentally different economic contract than a managed platform, and it deserves its own company, its own roadmap, and its own balance sheet.
Structuring them separately also makes each one more useful to customers. An organization can work with AutoGrant on grants without it being tied to a Microsoft platform engagement, and work with Centered Networks on its infrastructure without being sold grant software. The connection between the two is real and deliberate, but it is a connection between two clear, distinct things, not a bundle that obscures what either one does.
The next stage of AI is vertical.
For the last two years the story of AI has been the general assistant. One model, one chat box, ask it anything. That was the right place to start, and tools like Microsoft 365 Copilot are genuinely useful as a horizontal layer across email, documents, and meetings. But a general assistant is, by design, a generalist. It knows a little about everything and the specific rules of nothing.
The next stage is narrower and more capable. Instead of one model that answers any question, you give a domain to a team of agents that each know the rules, the language, the data sources, and the judgment of that field, and you wire them into the actual workflow of the work. That is what people mean by vertical AI, and it is where the real productivity is going to come from, because most professional work is not a single prompt. It is a sequence of specialized steps, each with its own standards and its own failure modes.
A general assistant knows a little about everything and the specific rules of nothing. The next stage gives a whole domain to a team of agents that each know their part of it cold.
Grant funding is almost a perfect example of why. Winning a grant is not one task. It is finding the right opportunity, reading eligibility rules precisely, judging whether you are actually competitive, extracting every required question from an RFP, writing a narrative that matches a funder’s priorities, building a compliant budget, submitting on time, and then reporting and renewing. Each of those is a different skill. A single chatbot will help with any of them and own none of them. A team of agents, each built for one stage, can own the whole arc.
What AutoGrant actually is.
AutoGrant is an AI grant operating system. Its premise is the opposite of most grant databases, which compete on the length of the list they hand you. AutoGrant’s position is simpler: most tools give you a longer list; AutoGrant tells you which grants you will actually win.
Underneath that is a team of specialized agents working across the full lifecycle, from discover and plan through write, submit, manage, report, and renew. In practice that means:
- Discovery that filters, not floods. The agents scan thousands of funding sources for genuinely mission-aligned opportunities, rather than returning every grant you are technically eligible for.
- An AI judge that scores your odds. Before anyone writes a word, an agent simulates a funder’s review committee and scores how competitive you really are, so a team spends its hours on the applications worth pursuing instead of long shots.
- Drafting in your own voice. The writing agents draft proposals in your organization’s voice, with citations and compliance checks tied to what the RFP actually asked for.
- Stewardship after the award. Submission, tracking, reporting, and renewal live in one place, because winning the grant is the start of the obligation, not the end of the work.
One design choice matters more than any feature. AutoGrant operates as a human-plus-AI partnership with approval gates wired in, not toggled on. You approve every consequential step, and AutoGrant never submits a proposal without explicit human sign-off. The agents do the heavy lifting of research and drafting; the people keep the judgment and the final word. That is the right shape for any serious vertical agent, and it is non-negotiable in a domain where your name goes on the application.
Why grants, and why this customer base.
Grants are the operating reality for the organizations Centered Networks has always served. Nonprofits, foundations, and rural hospitals run on funding they have to go out and win. For a rural hospital, a missed grant is a missed capability. For a nonprofit, it can be a program that does not run this year.
Time on the funder side at The San Francisco Foundation made that concrete in a specific way. The organizations losing winnable grants were not losing because their work was weak. They were losing to process — incomplete applications, narratives that didn’t track what the RFP asked for, budgets that didn’t map cleanly to a funder’s priorities. The gap is capacity, not quality. That is a solvable problem, and it is the one AutoGrant is built to solve.
It is also a domain where the cost of the old way is brutal and where the success-fee model genuinely changes the math. Organizations carry grants teams, pay consultants per application, or burn out the staff they have, and they pay for all of it whether or not a grant is awarded. A vertical agent that only charges when funding actually lands removes the budget risk from trying. For organizations whose missions are the whole point, lowering the cost and the risk of pursuing capital is not a convenience. It is access.
This also connects directly to work Centered Networks is already doing. There is, for example, real money on the table right now for rural hospitals through the Rural Health Transformation Program, where cybersecurity, IT modernization, and AI are named, allowable uses. Centered Networks can scope and price the technical work; a tool like AutoGrant is built for the work of actually finding and winning the funding that pays for it. Same organizations, two different problems, two companies built to solve them.
How the two companies reinforce each other.
Separate does not mean disconnected. The two businesses share a worldview that I care about: that AI should be governed, accountable, and pointed at the missions of organizations that do not usually get to go first with new technology.
Centered Networks is the platform side of that worldview. It makes sure the environment an organization runs on is secure, compliant, and ready for AI, so that when agents touch real data, they do it inside guardrails a board can stand behind. AutoGrant is the workload side. It is a concrete, high-value example of vertical agents doing real work for exactly the kind of organization Centered Networks serves. The governance Centered Networks builds is what makes vertical AI safe to adopt; the vertical AI is what makes the governance worth the effort.
Founding both lets me build each one properly. Centered Networks stays focused on being the calm, credible partner for the Microsoft platform. AutoGrant stays focused on being the best system in the world for finding and winning grants. And the through line, the thing that made me want to build them at the same time, is the conviction that the next decade of AI belongs to specialized agents that know one domain deeply, deployed for people doing work that matters.
I am building two companies because they are two businesses, and I am building them now because the shape of useful AI is changing fast. The general assistant was the opening act. The vertical agent, built for a domain and accountable to a person, is what comes next. AutoGrant is my version of that bet for the part of the economy I care most about, and Centered Networks is the foundation that makes it safe to stand on.
If you run a nonprofit, a foundation, or a rural hospital, both of these are built for you, from two directions, by one team with a single conviction: that the organizations doing the most important work should be the ones who go first with technology that actually works.
See AutoGrant.
An AI grant operating system that tells you which grants you will actually win, drafts them in your voice, and only charges when funding lands. Built for nonprofits, foundations, and the missions that run on grants.
Visit autogrant.aiGet your platform AI-ready.
Before agents touch real data, the platform underneath has to be governed. A Discovery Sprint maps where you are and what it takes to adopt AI your board can stand behind, in two weeks.
Start a Discovery SprintAutoGrant and Centered Networks are separate companies founded by Chris Grecsek.