Proposal for Gleave Partnership · 21 July 2026

Hellovia

A project's status is scattered across Google Chat, Gmail, Buganizer, and Sheets, where decisions, risks, and promises scroll past and get lost. Hellovia reads every project signal, keeps a living summary, and turns what matters into suggested work items. A person stays in charge of every decision.

A working demonstration already reads real project conversations.

The project already talks. Nobody can listen to all of it.

Our programmes run across Google tools: vendors confirm deliveries in Chat, decisions land in Gmail threads, defects move through Buganizer, and trackers live in Sheets. A programme manager has to read all of it, remember all of it, and retype what matters between systems by hand. Things slip, not because people are careless but because the volume is inhuman.

Risks surface late. A delay mentioned in passing on Tuesday becomes a crisis two weeks later.
Actions get buried. “Can you follow up with the vendor?” disappears into scroll-back within hours.
Hours are lost to retyping. Reading, summarizing, and copying from one system into another is a daily manual job.

Every signal, one pipeline

We propose one pipeline for every project signal. The machinery is source-agnostic: Gmail, Buganizer, and Google Sheets plug into the same path from capture to review, and each new source is a connector, not a rebuild. The Google Chat connector is built first as the working demonstration.

Google ChatDemonstrated
GmailProposed
BuganizerProposed
Google SheetsProposed
Suggested work itemsawaiting review
Living summariesevery claim cited
Registers & reportsofficial records, digests in Chat

Five steps, one human decision

The team works as usualEveryone keeps working where they already work. Nothing new to install or learn; Google Chat is connected first.
Hellovia capturesEvery signal from a connected source is recorded the moment it happens, including edits and deletions.
The AI reads in contextIt knows the project, the surrounding history, and what has already been raised, so it doesn't repeat itself.
It draftsSuggested actions, risks, and questions, plus a living summary of each conversation. Every claim cites the exact message it came from.
You decideEverything is a draft until a person approves it. The AI never creates an official record on its own.

Proof first: a working demonstration

The Google Chat connector already runs for demonstration purposes, reading real project conversations end to end: capture, living summaries with citations, suggested work items, and a review screen for each. Its AI brief is written from the real Mechanical Programme Manager job scope. Below, its unedited output from a single message.

LC
Leonard Chan · project space, 17 July
“The CRAH unit delivery slipped to next Friday”
what Hellovia produced
Issue
CRAH unit delivery schedule slip
Priority high, risk high, confidence 1.00
Information request
Request CRAH tracking details and impact assessment
Priority high, confidence 0.95
Risk summary
Delivery slipped to next Friday; may impact mechanical installation and commissioning milestones37
One line of the conversation's living summary; the chip links to the source message

Produced by the demonstration system during acceptance testing, not a mock-up. The summary also correctly refused to mark the delay “resolved” when a later, ambiguous “all good now” message arrived: it preserves uncertainty instead of guessing.

What the demonstration changes

A programme manager stops reading every thread, holding it all in their head, and retyping what matters. The job becomes reviewing a short list of drafted items and deciding what is real, with the source one click away.

The guardrails are the product

A human approves everything

The AI suggests; it never files, assigns, or sends anything on its own. Official records only exist after a person says yes.

Every claim is cited

Each summary line and suggestion links to the exact message behind it. No floating AI opinions; you can always check the source.

Data stays in our cloud

The AI runs inside our own Google Cloud project. Project data never leaves the Google ecosystem and is never used to train models.

Scoped by policy

Work-item suggestions only run on conversations a person has assigned to a project. Summary coverage is a per-organization setting.

Spending is metered

Every AI call is measured and logged, with a daily budget switch that pauses spending without losing work. At demonstration volume: well under a dollar a day.

Deletions are honored

When a message is edited or deleted at its source, Hellovia updates or removes it everywhere. Content not assigned to any project expires automatically after a retention window.

Who gets in, and what gets recorded

Hellovia is closed by default. Access exists only by invitation, every session is verified and revocable, and every administrative action lands in an audit trail. Walk the path a person takes:

No self-signup. Every account starts as an invitation.

An administrator invites each person by name into a specific role. Invitations expire on their own, can be revoked before they are accepted, and the invitation list is visible at all times.

Where this goes next

Demonstrated
  • Full chat capture across the organization
  • Living summaries with citations, per conversation
  • AI-suggested work items awaiting review
  • Health & cost monitoring
Later
  • Reports back into Chat: daily and weekly digests where the team already is
  • Ask the project: question-and-answer over the full history; the groundwork is already built
  • More sources: Gmail, Buganizer, and Google Sheets through the same pipeline