Pinley: launch readiness and the Aug/Sep plan¶
For Conor. From the dev team. 2026-08-06.
Conor, we sat down as a whole team on Aug 4 for a long session to get everyone on the same page. Everyone read your prototype and the Cloud Design, then we went through the live app module by module and lined the three up against each other. Here's what came out of it: how we see the app, what's still in the way of the first launch, the plan for August and September, and the handful of decisions we need from you.
How we see the app¶
It's one line from an incoming bid to a running project:
- Bid invites come in from BuildingConnected, from manual entry, and (through Smart Import) from SharePoint. On the invite you see the scope, the timeline, and the document list. We only pull the documents into our storage once you accept the bid.
- Bid detail: source, status, priority, financials (hidden by role), an AI-shortened version of BC's long scope text, and competitors (historical).
- Estimate: sections, scope lines, sub-lines, vendor quotes, buy vs bid, markup to a contract price. Per-GC revisions you can compare. Numbers go in by hand today, and vendor RFP is still manual (email, then copy the numbers back in).
- Takeoff: the estimator marks systems on the drawing and those marks become estimate lines. The AI auto-takeoff on vectorized PDFs works, but it's a helper right now. The real work is still manual.
- Proposal / SOV: a clean PDF (lump sum, on letterhead) for an awarded bid, shown by scope rather than the full breakdown.
- Award to Project: when you set a bid to Awarded (a call you make off the platform), we create the project and its folders, and the final estimate becomes the budget.
- Running the project: an AI handoff brief, the team, spec sections, budget and change tracking, the controls (Submittals, RFIs, Change Orders, each with an AI draft), documents in role-gated folders, drawings for progress, and Daily Logs on mobile.
- Cara and Smart Import: Cara is the AI copilot over your data (it drafts, you approve). Smart Import is the single-login SharePoint browse, review, and commit flow for bringing existing data in.
The short version on the build: the core above is shipped and holding up. Roughly 85% of the value path works today.
The three sources of truth, and which one wins¶
| Source | Date | What it's the source for |
|---|---|---|
| Prototype | ~June 2026 | the shipped core (bid, estimate, award, project, daily logs, Cara, Smart Import) |
| Cloud Design | 2026-07-27 | the look, and the bigger modules coming (Executive, Takeoff Engine, Progress Drawings, floating Cara) |
| The live app (dev/staging) | now | what's actually built, about 85% of the core |
Where we land: for the shipped core, the prototype is still the reference, and we shouldn't rebuild what already works. The Cloud Design is the right target for the look and for the bigger modules coming (Executive, Takeoff Engine, Progress Drawings), and we'll do those as their own wave, not as a launch blocker. One thing we want to say plainly: the Cloud Design isn't a spec for the shipped core. It moves fast, some of it is weeks stale, and pages get duplicated every time it regenerates. Treating it as the record of what's "done" is a big part of why the timelines feel off. Our fix is in the last section: one place that's the source of truth, and a simple gate every new idea passes before it becomes work.
The first launch¶
The first launch is a quiet internal pilot for two people, Jim (estimator) and Alex (PM). The scope is the core bid-to-project flow, plus importing their existing bids and projects.
What's out for launch (manual is fine for now, or it waits): automatic vendor RFP emails, the fancy Daily-Log drawings, voice-to-text, the deep Project-Drawings progress module, compliance and insurance, and outbound bidding.
What "it worked" means: Jim and Alex import their real data with no corruption, and run a bid all the way into a project without losing anything.
The one thing that matters most is Smart Import being solid. Jim and Alex will build estimates on top of imported data, so if we ever had to wipe the database to clear corruption, we'd wipe their work and their trust along with it. It needs to be near perfect before they touch it, and we need to prove a BuildingConnected bid all the way through (documents, takeoff, budget). We're close, and QA is on it now.
What's still in the way¶
Buckets: Blocker (fix before Jim and Alex), Launch (should be in the pilot), then August, September, and later. Sizes are S/M/L/XL.
Blocker and launch¶
| Gap | Bucket | Owner | In the pilot? | Size |
|---|---|---|---|---|
| Smart Import data corruption | Blocker | Ivan B | yes, this is the gate | L, in progress |
| BuildingConnected bid not yet proven end to end (docs, takeoff, budget) | Blocker | Marijana + Artem | yes | M |
| Proposal template doesn't match yours (need your real one, show by scope only) | Launch | Ivan B | yes | M-L |
| Vendor RFP emails off by default for the pilot (allowlist + a test vendor) | Launch | Marko + Artem | yes | S-M |
| Confirmation prompt on every delete (a misclick loses data today) | Launch | Frontend | yes | S-M |
| Buttons that look disabled but still click | Launch | Frontend | yes | S |
| AI takeoff symbol overlay lands in the wrong spot | Launch | Artem | yes | M |
| "failed to fetch" errors (add retries, watch the cluster) | Launch | Marko | yes | M |
| Make sure PM / assistant PM can't see margin or profit | Launch | Marko | need to verify | S-M |
| The bid-invite "Acknowledge" step isn't wired up | Launch | Product | need your input | S |
August, right after launch¶
Vendor portal flow (the vendor sees too much today), competitors field showing only after a loss plus a BC source, bid-invite expiration alerts in QA, self-hosted error monitoring, PO approval retest, the "applying a template creates an empty revision" bug, matching the look between bids and projects, a docs/drawings map, automated RFP with reply parsing, and better handling of long AI operations (progress plus a notification when it's done).
Already moving, not waiting¶
Daily Logs new design is largely built and in QA. The interactive progress-drawing piece is being built right now by Artem, and he'll bring you two options to pick from (more below). Takeoff Engine: the base AI auto-takeoff already ships, and the accuracy and AI-first push is the top thing after that.
Later¶
Compliance and insurance, lead vs junior estimator permissions, outbound bids, the full self-hosted takeoff model, and the Executive suite (Dashboard, Master CO Log, WIP, Bid Targeting) once we have the specs and the accounting hookup.
The August/September plan¶
One thing to set up front: a real feature now takes multiple cycles (4 weeks or more), because it goes through QA and comes back a few times before it's actually done. That's the price of it being stable, and it only holds if we don't change the scope mid-cycle.
| Wave | What | When |
|---|---|---|
| Launch | the blocker and launch items above, Smart Import and a BC bid proven end to end | now, we launch on QA sign-off |
| August | finish Daily Logs (in QA plus the progress drawing), then stabilize: vendor flow, PO retest, error monitoring, matching look, long-AI handling | August |
| September | Takeoff Engine first, then the other Cloud-Design modules | September |
| Later | accounting integration, live Bid Targeting signals, the full self-hosted takeoff model, per-tenant fields | after the pilot |
Your top two, the way we understand them:
- Takeoff Engine, AI measures first and the estimator verifies. This is your number one going forward. The base pipeline (scale read, system detection, symbol groups, the training tab) already ships. The September work is accuracy and the measure-first-then-verify flow. The priority is set, but how fast it moves comes down to two things from you: a clean set of drawings we can train and test on, and a yes on the self-hosted vision-model approach.
- Daily Logs new design. High priority, and finishing now. The design is built and in QA. The interactive progress-drawing part is being built today, and Artem will bring you A or B to pick. A is mark progress inside the daily log (lighter). B is the full interactive markup (heavier). Voice stays out for now.
The rest of the forward wave, each waiting on a decision or a dependency:
| Module | Where it is | Size | Waiting on / decision |
|---|---|---|---|
| New look (navy/copper, Nunito + DM Mono) | partly done | M-L | pilot, or after? |
| Floating Cara button | done | - | shipped |
| Executive Dashboard (empty state first) | stub | L | turn it back on? plus the accounting hookup |
| Master CO Log | not built | L | spec |
| WIP report | not built | M | accounting job-cost sync |
| Bid Targeting | stub | XL | spec, plus outside data feeds |
| Compliance (COI, rules, renewals) | not built | L+ | timing |
Inside the Executive suite we'd do the Dashboard (empty state) and Master CO Log first, since they're the least blocked.
What we need from you¶
- Proposal template: send us the real one you use, the rebuild depends on it. And confirm we show the final by scope, not the full breakdown.
- Takeoff Engine (your #1): the direction is set, AI measures and you verify. To move fast we need the clean drawing set and a yes on the self-hosted model approach. Those two set the pace.
- Daily Logs progress drawing: pick A or B, Artem brings both today. A is mark progress in the log. B is the full interactive markup. We lock it and finish.
- The Acknowledge / "intend to bid" step: how should it work? We might not have the GC's email, or we may need to flip an "intend to bid" flag in BuildingConnected.
- The new look (navy/copper): do you want it for the pilot, or after? Our take: after, it touches every screen.
- Compliance: September, or later? Our take: September, after Takeoff Engine.
- Competitors data: where should it come from (BC webhooks?), and should it only show after a bid is lost? Our take: only after a loss.
- Per-tenant custom fields (for white label): when do you want this? Not a launch blocker.
- Lead vs junior estimator, and who sees margin: pilot, or later? Our take: later, but we'll confirm no margin leaks at launch.
- The blocked ones: we need direction on the Bid Targeting and Master CO Log specs, whether to turn the Executive Dashboard back on, the Azure Sites.Selected permission (that's the real security boundary for Smart Import, and it's a change on your Azure side), and the accounting integration timing: pull it into September, or ship Executive/WIP with honest "no live numbers yet" states until it's ready?
How we'd like to handle new ideas from here¶
The thing under most of the friction: a new idea (usually a Claude-design screen) shows up mid-cycle and gets treated as already agreed, which knocks over whatever's in QA and pushes the date. Two changes fix it.
One source of truth. The roadmap (the Feature Matrix) is the one place scope, priority, and status live. Not the Cloud Design, not Slack, not a doc that keeps moving. The Cloud Design stays the picture of how it should look, it's just not the spec or the status. If it's not in the matrix with an owner, a priority, and a status, it's not committed.
A simple gate: capture, review, lock, build.
- Capture: the idea goes in one place (a Linear issue with a link to the design screen). Just captured.
- Review: we scope it, what it touches, what it depends on, rough size.
- Lock: you sign off on scope, priority, and which wave. Once it's locked it doesn't change mid-cycle. New thoughts go to the next planning point.
- Build: only locked things go into a cycle, and "done" means it cleared QA, not that the first PR merged.
We'd also swap the standing weekly call for topical deep-dives: the right person in the room, an agenda sent ahead, only when there's a real decision. That's what lets us give you dates we can actually hit.
Bottom line¶
The core is built and steady. Daily Logs (new design) is finishing in QA now. The Takeoff Engine base already ships, and the AI-first accuracy push is your number one going forward, its speed comes down to the drawing set and your yes on the model approach. The launch itself is waiting on Smart Import being solid, and we're close. The one change that makes the dates reliable is the source of truth plus the capture, review, lock, build gate. The decisions above unblock the rest of August and September.