Deloitte Launches Edisyl to Sell Expert Knowledge as an AI Asset

Deloitte introduced Edisyl on November 12, 2026, as a Boston-based enterprise AI venture whose core product packages 20-year practitioner judgment and proprietary company data into portable "Knowledge Packs" that work with Claude, Codex, Gemini and Copilot. The new company emerged from Deloitte's July 2026 acquisition of Flipside Crypto's AI business, a deal that brought the entire 24-person operating team of that unit into the new venture.

The launch matters because it targets a specific and expensive gap in enterprise AI deployment: organizations that already run AI assistants and agents but cannot get business-specific answers from them without extensive custom engineering. Edisyl's answer is to treat institutional and expert knowledge as a managed asset — with owners, versions and maintenance costs — rather than as something rebuilt for every new model.

What exactly is Edisyl selling?

The product has two components. Practicing experts annotate the Domain Pack, a curated collection of field-specific material that Edisyl updates on an ongoing basis. The Proprietary Pack contains an organization's own terminology, definitions, policies and data, which remain under the organization's control.

Deloitte says the combination lets existing AI systems handle business-specific questions with supporting sources and professional judgment attached. Edisyl's product site states the external component can draw on:

• partner and market data
• government and regulatory material
• standards and codes

To assemble and update those resources, the company uses agent fleets — coordinated sets of AI agents performing defined tasks — with subject-matter experts directing their work. Under both components, a semantic layer (a mapping between structured system terminology and human language) translates the terms in company systems into the words employees actually use.

Who owns the knowledge, and where does it run?

Deployment is customer-owned. Edisyl says customers can install a Knowledge Pack in their own cloud or on their premises and connect it to Claude, Codex, Gemini or Copilot. Both halves of the pack are constructed and installed as a unit, and ownership is retained by the customer.

Portability is the central design decision. A pack is intended to move between AI systems rather than require a fresh build for each tool. The stated goal is a knowledge layer that outlives the model underneath it, with context managed as an asset that has owners, versions and ongoing maintenance costs. That framing positions Edisyl not as another model vendor but as infrastructure for the layer above the model.

The first packs are still in development. Edisyl is building them with design partners and named domain experts, including Deloitte programs and Fortune 500 organizations across industries. That places the initial product work with organizations directly participating in the construction of field-specific knowledge bases, rather than with general availability customers.

Where did Edisyl come from?

The venture sits within D71 Ventures, Deloitte's portfolio for AI-native businesses, which supplies staged funding and hands-on company-building support alongside Deloitte's industry reach to create standalone ventures. Edisyl is structured as one of those standalone companies, not an internal Deloitte product line.

Dave Balter, previously Flipside's chief executive, continues as an Edisyl co-founder. His leadership team includes:

• Kyle Fugere as chief operating officer
• Jim Myers as chief technology officer
• Andrew "Mo" Septimus as chief financial officer
• Eric Stone as chief data scientist
• Gabriela Serret-Campos as senior vice president of People Success

The July 2026 transaction did not include Flipside's crypto-data business, which Sonaverse Inc. bought separately. Balter's earlier ventures include BzzAgent and Smarterer, acquired by dunnhumby and Pluralsight respectively, among the eight companies he has founded. He subsequently led mergers and acquisitions at both acquirers.

Why does the 'annotation layer' argument matter?

Balter tied the product directly to expertise that sits outside conventional corporate data. "What AI tools can't access is the judgment of the person who has worked a specific problem for 20 years. That's what we set out to package, and it's why the annotation layer matters just as much as, if not more than, the interface," he said.

The claim is a bet on where enterprise AI value concentrates. If Balter is right, the durable asset is not the model — vendors will keep replacing those — but the curated, expert-annotated context layer that any model can query. That inverts the usual procurement logic, in which companies invest in a single assistant platform and rebuild integrations when they switch.

What happens next?

The design-partner work now connects the acquisition to the product strategy: Edisyl's central task is to maintain expert annotation and customer-owned context as a reusable enterprise asset while the AI systems around it change. Until the first packs ship beyond the design-partner cohort, the venture's proof point will be whether Fortune 500 participants and Deloitte programs can turn the two-pack structure into knowledge bases that genuinely transfer across Claude, Gemini, Codex and Copilot without a rebuild — the portability promise on which the whole model depends.