SAN FRANCISCO – September 23, 2026 – Jinba, the YC-backed AI workflow platform for regulated enterprises, has introduced a McKinsey slide deck skill inside Jinba App — a capability that turns generic, obviously-AI-generated slides into a consulting slide deck an executive team can present to the board.
The skill is authored by a former McKinsey consultant who went on to co-found the company, and it encodes roughly 110 slide-design rules collected from real consulting review feedback into a rulebook the AI reads before producing every deck. That is the distinction that makes the output a genuine consulting slide deck rather than merely a polished one: the open-source version of the skill attracted 461 GitHub stars and 41 forks, and an accompanying writeup on the company’s official Note account drew more than 260 likes.
Most AI presentation tools produce decks that look clean but collapse on closer reading — titles that open with category labels like “Current State and Challenges,” three-column layouts that offload the thinking onto the reader, and a conclusion repeated three times across title, body, and a bottom banner. Jinba’s thesis is that the difference between a generic AI deck and a consulting slide deck is not a nicer template — it is a set of review rules that survive from one AI session to the next.
The skill ships as part of Jinba App rather than as a standalone tool, so teams generate decks inside the same governed, permission-controlled environment they already use for workflow execution. The open-source rulebook and parts library are available on GitHub for the developer community, while the managed, team-governed version runs inside Jinba App.
Key Capabilities
– A 110-item slide-design rulebook — the “source of truth” file encoding assertion-first titles, premise-to-conclusion layouts, and consistent terminology, accumulated from real review feedback.- A 62-type slide catalog — a visual reference of proven consulting layouts, from issue trees to delta bridges to executive summaries, used as an idea book rather than a rigid template.- Machine rule checking — an automated checker (`check_deck.py`) that flags rule violations, leftover placeholders, and structural defects before a human ever sees the deck.- Independent review — a built-in practice of having a separate AI agent that did not build the deck read it back for logic, contradictions, and language, mirroring a real consulting review pass.
About the Approach
The skill’s public release is part of a broader pattern for how Jinba builds capabilities: encode professional judgment into a structured, persistent format that AI reads each time, rather than relying on ad-hoc prompting. This mirrors the company’s “Company Brain” direction, where enterprise knowledge is modeled into a bespoke schema instead of retrieved by surface similarity — the same reason generative AI pilots fail in banking and insurance when output looks polished but isn’t grounded in the right context.
The technique has already been validated against leading models. The company published side-by-side comparisons of Claude Opus 5 and GPT-6 Astra on identical deck-building tasks, showing the skill upgrading output quality on both — an extension of the same AI governance discipline Jinba
About Jinba
Jinba is a San Francisco-headquartered, YC-backed (W26) AI workflow platform that lets business users in highly regulated industries — banking, insurance, legal, healthcare, and pharma — build production-ready AI workflows with natural language prompts that compile directly into APIs and MCP servers. Its SOC II–compliant platform ships governed automations in days rather than months, on-premise or air-gapped, with deterministic execution, audit logging, RBAC, and SSO. For more information, visit jinba.io.
Media Contact: Jinba Media Relations [email protected]
Media ContactCompany Name: JinbaContact Person: Media RelationsEmail: Send EmailAddress:1301 16th St City: San FranciscoState: CACountry: United StatesWebsite: https://jinba.io
