The plan we’re testing
Reuse first. Review where it counts.
Piko starts from what your firm already knows and puts a front door in front of it: a brief you can act on, evidence you can check, and review reserved for the moments the business adopts a position.
01Reuse
02Change awareness
03Durable memory
Why reuse comes first
The review queue is real. It shouldn’t be the front door.
Today, Piko earns trust with an evidence-checked review queue: every claim carries a verified quote, and nothing counts as approved knowledge until a person signs off. That’s the right foundation, and it stays.
But if the first thing a new user sees is a queue of things to approve, Piko looks like paperwork before anyone’s felt the relief of finding something they thought was lost. Piko puts reuse first and keeps the review queue for what actually deserves it: a position the business is about to stand behind.
Being honest about the market
This isn’t an empty category.
Organizing information by meaning instead of location, connecting existing tools without a migration, maintaining knowledge with human review, running AI on hardware you control: each of these already exists somewhere, in wikis with automated maintenance, in connected-search tools, in self-hosted AI stacks built for exactly this kind of use.
No single piece of this is undiscovered. The combination is the point: reuse, change tracking, and infrastructure you control, in one system that keeps its receipts.
How we’re sequencing it
Reuse first. Change awareness next. Durable memory after that.
Three connected ideas, tested in order, each earning the next.
First
Reusable project memory
The Living Brief becomes the front door: a working brief for the project in front of you, not a search box that only helps if you already know what to type. The first release has to make one real, messy archive genuinely useful before anything else changes.
Then
Change-aware memory
A reissued source is re-anchored, and every decision that load-bearingly depends on the changed claim moves to review automatically — in the originating project and in any other workspace that relies on it. Recorded dependencies only; never a claim to catch every consequence.
Later
Durable memory for AI-assisted work
As producing a first draft gets cheaper, saving another report becomes a weaker way to preserve its value. A new research run first consults prior work, identifies what changed, keeps the conclusions that still hold, and isolates the questions that are genuinely open. Research accumulates instead of piling up.
Who we’re targeting, and who we’re not
A specific kind of firm, not a company size
Built for specialist systems integrators and technical consultancies of roughly 10 to 75 people: repeat work, changing requirements, years of files, and one or two people who act as the company’s memory.
- Not a business whose main problem is accounting, scheduling, or inventory, rather than knowledge.
- Not a business already satisfied with a well-governed suite it has fully adopted.
- Not a deployment that needs a major compliance program before anyone has seen whether this helps.
- Not a request to preserve one person’s knowledge when it was never written down anywhere. Piko can’t reconstruct reasoning that doesn’t exist in a document. It needs a low-friction way to capture it going forward, through brief explanations or decision confirmations, not a promise to recover the past.
The whole picture
What Piko does
Every capability below is marked plainly. Nothing on this page states a future capability as something you could use now.
| Capability | Status |
|---|---|
| Every claim carries a quote checked verbatim against its source before it’s stored | |
| A person approves, rejects, or returns every claim, tied to a real, authenticated session | |
| Decisions are records with an audit trail, not memories | |
| A changed source is re-anchored, and decisions that load-bearingly depend on it move to review, ranked by commitment date. Decisions that merely mention it stay silent, on purpose | |
| Local retrieval and the review queue keep working without an internet connection | |
| No per-token model fee for local inference | |
| A Living Brief that assembles relevant precedent, decisions, and open questions for a project automatically | |
| Discovery — indexing, connecting, drafting a summary — that works without approving each step | |
| Read-only connectors to shared drives and one cloud ecosystem | |
| A distinction between a firm’s working synthesis and its accepted position, so a fresh draft never quietly overwrites what’s already agreed | |
| Read-only access for outside AI assistants through MCP, scoped the same way a person’s access would be | |
| Cloud model providers and paid web search |
Scope, on purpose
What we’re deliberately not building yet
- No new file server. Files stay exactly where they are.
- No automatic reorganizing of a shared drive.
- No general-purpose agent that acts on the business’s behalf.
- No unlimited historical cleanup bundled into onboarding.
- No hosted or multi-company version. Piko still runs on hardware the business owns.
How we’ll know if it’s working
Signals we’re watching, not settled conclusions
- Whether a brief is useful on a messy real archive, without heavy cleanup first.
- Whether the people using it spend less end-to-end time on a real recurring task, review included.
- Whether the changes it flags are ones people agree actually mattered.
- Whether a firm keeps paying for it once the novelty wears off.
These are proposed signals, not settled thresholds. If the strongest value turns out to live in one narrow workflow, we’d rather narrow around it than broaden the pitch to fit.
Tell us if this matches your firm
This is a plan we’re testing, not a finished roadmap. If the reuse problem above sounds familiar, we’d like to hear about it.