ibl.ai and the AI You Can Fire

The student wanted Fridays back. In Syracuse University’s published account of Clementine, its AI class-search assistant, the request was wonderfully plain: a workable schedule, with certain hours ruled out. No manifesto about artificial intelligence. Just the ancient undergraduate ambition of arranging a week one could bear to live through.

That is a useful place to begin with ibl.ai. The company sells software with a grand name, an AI operating system. Its value is easier to understand in a small moment: a student asking a question in ordinary language, an agent consulting institutional information, and a university deciding how that exchange should work.

The right to leave, sold at the entrance

Most buyers understand a software subscription. Fewer have had reason to ask what happens when the subscription ends. Does the application stop? Can the integrations survive? Who has the code? These sound like questions for the lawyers until a system becomes something people rely on every day.

ibl.ai puts those questions near the beginning of the sale. An organization can use its hosted service or arrange a deployment on infrastructure it controls. The full codebase is offered under a perpetual license through a custom-quoted engagement. The attraction is straightforward: the supplier can help build the system without remaining the only party allowed to change it.

Syracuse supplies a concrete example. The university says it received the platform’s source code and runs it on its own Google Cloud, using Vertex AI. Identity and role-based access connect it to campus life. The institution can choose models and develop its own interface and applications. A software purchase becomes a foundation its staff can work on.

“We wanted full ownership of our AI platform - the code, the data, the infrastructure. IBL.ai delivered exactly that.”Jeff Rubin · Syracuse University

The distinction matters in procurement. Keeping a copy of your documents is useful. Keeping the application that makes those documents useful is another kind of control. The buyer still needs a team capable of running it. A source-code handoff transfers options and obligations together.

The meter, the seat and the invoice

Imagine an institution offering AI access to 10,000 people. At a hypothetical $20 per person per month, the seat bill alone is $2.4 million a year. That calculation says nothing about how often anybody uses the service. A person who asks one question and a person who asks a thousand occupy one seat each.

ibl.ai offers a different purchasing shape. Its self-serve route uses prepaid credits. Its advertised enterprise pilot starts at $15,000, with one or two agents and one integration; the fee credits toward a fuller engagement. Integration and deployment is listed at $25,000 to $80,000, one time. Full source-code transfer requires a separate quote.

For a self-hosted system, the continuing bill includes model usage and infrastructure. Maintenance, security work and internal staff time also belong in a buyer’s calculation. “No per-seat fee” tells you how the vendor charges. It does not tell you the total cost of operating the service.

Syracuse reports roughly 85% lower costs against its per-seat comparison. That is an encouraging account of a particular deployment, not a discount every customer should put into a spreadsheet. Usage intensity, model selection and the work needed to connect existing systems can all change the arithmetic.

A catalog is not a degree audit

Clementine Class Search launched in spring 2026. During its first 37 days, 2,038 students asked 8,217 questions across 2,401 conversations. Thirty-seven percent of activity happened outside weekday business hours. There is a practical constituency for software that stays awake after the advising office closes.

The university also describes the boundary students ran into. They wanted to know whether a course satisfied their requirements and what they should take next. The agent knew the catalog. Personalized degree advice would require a connection to each student’s degree audit.

This is a more instructive limitation than a spectacular chatbot blunder. An answer can be articulate and still lack the record that makes it relevant. “When does this class meet?” and “Will this class help me graduate?” may sound adjacent. The second question needs a different evidentiary foundation.

Students’ questions gave Syracuse a reason to expand the scope. The university identifies student-aware degree-audit integration as the next step. Owning the platform gives it a way to pursue that work; ownership alone cannot supply the missing data.

Six products, one set of institutional plumbing

The core Agentic OS connects organizational systems to agents. Around it sit Agentic Vibe for application development, Agentic LMS for learning and skills, Agentic Course for course creation, Agentic Video for avatar and video production, and Agentic Wallet for verifiable credentials. The lineup retains a strong educational accent even as the company markets to government, healthcare, legal, financial-services and enterprise buyers.

Those buyers share a familiar problem. Information sits in different systems, with different permissions and different meanings. An agent helping a learner may need course content and enrollment information. An administrative agent may need a policy document and an operational record. Connecting everything indiscriminately would be a poor substitute for connecting the right things.

ibl.ai’s engineering service builds connectors using the Model Context Protocol, or MCP, with role-aware access, logging and policy controls. Customers retain connector code, deployment scripts, schemas and runbooks. That is where much of the expertise resides: making an existing organization legible to an agent without flattening its rules.

For a faculty member, the work can begin more simply. Syracuse’s guide describes uploading approved teaching materials, choosing a model, setting instructions and controlling who sees the agent. It cautions that text-based sources work best; scanned PDFs and image-heavy slides provide less usable material. Even the humble document format gets a vote.

The family behind the handoff

ibl.ai is family-owned and operated. Mikel Amigot is its founder and CEO; Miguel Amigot leads technology, Blanca Amigot finance, and Jaione Amigot engineering operations. Mikel also founded IBL News and publishes a book about the AI-native university. Education is a recurring occupation here, rather than merely an industry added to a sales dropdown.

The company’s earlier educational software work used Open edX. Its older account describes building a platform on that foundation for universities and corporate training customers. Today, the emphasis has moved toward agent infrastructure and institutional ownership. The continuity is the engineering around learning systems; the new surface is conversational and increasingly capable of taking action.

Within the market, ibl.ai sits between a ready-made assistant and a completely internal build. General assistants, enterprise knowledge platforms and an in-house stack are all plausible alternatives. Syracuse itself offers Clementine alongside Claude Enterprise, Copilot and Gemini. The tools can coexist because a broad assistant and a campus-specific agent need not earn their keep in the same way.

A pilot small enough to judge

The most transferable idea is to give the pilot a narrow job. Choose a workflow with accessible data and an identifiable owner. Define a useful answer, an unacceptable action and the point at which a human must take over. Then test with the people who will actually use it.

Measure more than conversation counts. Check whether the answer is correct, whether permissions hold, what each completed task costs and whether staff must repair the result. ibl.ai’s published release notes describe continuing work on spend limits, privacy filtering and audit visibility. Those are operating concerns that persist after the first successful demonstration.

This approach has conditions. An organization unable to maintain infrastructure may prefer managed hosting. A team without usable records will spend time preparing them. An occasional user may have little reason to commission a custom deployment. And a high-stakes decision still needs accountable human judgment, however politely the agent speaks.

What the Clementine story makes visible is a sequence worth copying: start with a real inconvenience, observe the questions people ask, and let the next integration answer an observed need. ibl.ai’s ownership pitch gives institutions room to do that work themselves. The value of keeping the code is what they can choose to do with it tomorrow.