Wyecliff: TimeHarmony
How we built our own AI-assisted time and budget system, so our team barely touches timesheets and clients never pay for the paperwork.
Background
Wyecliff is an AI implementation firm. A lot of what we do for clients is turn scattered tools and manual work into a single source of truth, so we hold ourselves to the same standard.
This one had a specific belief behind it: clients should not be paying billable hours for our team to fill out timesheets, and our people should not be losing valuable time to time tracking either. But our own time and project economics lived across a few systems plus whatever people remembered to enter by hand. There was no reliable way to see billable versus admin time, watch project budgets in real time, or catch overruns before invoicing. Managers could not see unsubmitted time, and Friday project reporting meant stitching numbers together instead of reading one dashboard.
So we built the thing we would have built for a client: one authenticated system where time is captured as you work, reviewed against estimate phases, and rolled up against project budgets, with AI doing the parts that used to eat people's time.
What We Did
We built TimeHarmony: a web app plus a Chrome extension that turns everyday browser activity into phase-correct, submit-ready time.
- Capture as you work. The extension passively tracks activity, domain and page title only, never keystrokes or screen contents, so logging time takes no real effort. It requires a project and phase before tracking starts, so there is no unassigned time, and AI infers the right project and phase from what you are actually doing, even suggesting a switch when the work changes.
- Review in minutes, not hours. A timeline lays out the week in simple blocks you can adjust, and a review view rolls everything up by project and phase so submitting time is a quick pass rather than a chore. AI writes plain-language summaries of what happened in a day, week, or month, so the notes write themselves.
- Budgets in real time. Every project carries its estimated phases and budgeted hours, and a live budget view shows how actual time is tracking against the estimate, including hours logged but not yet submitted. Overruns show up in the middle of an engagement instead of after the invoice, and managers can review and approve time from the same place.
- The whole thing runs on the cadence we actually operate on: time in by end of week, review and approval, then monthly close and invoice handoff.
- Under the hood. It is a production build, not a prototype: a React and TypeScript web app, a Chrome extension, and a secure cloud backend, with AI running behind authenticated, rate-limited calls and scheduled syncs that keep it reconciled with our accounting and timesheet systems. Claude handles the project inference, and a second model handles the written summaries.
Impact
The weekly time and budget review that used to span several systems and a fair amount of memory now happens in one place. The team captures time as they work and cleans it up in minutes, managers see pending versus submitted hours and approve without hopping between tools, and leadership can watch budget against estimate mid-engagement instead of after the invoice. Because the phase structure matches the way work was scoped and sold, submitted time lines up with the estimate.
It solved the problem we set out to solve: our people spend almost no time on timesheets, and clients are not paying for us to do administrative work. It is also proof of our approach. Consolidating capture, AI, and back-office systems into one loop is exactly the pattern we build for clients, so when we tell a client we can replace a brittle timesheet stack with something that tracks, reviews, and reconciles to their billing tools, we are describing something we already run every week. See also our Marketing Command Center and custom AI builds.
Weekly time review instead of hours across tools and memory
Passive capture with project and phase required up front
Budget versus estimate visible before the invoice
Clients are not paying for timesheet paperwork
Software: React, TypeScript, Vite, Tailwind, Supabase (Postgres, Auth, RLS, Edge Functions), Chrome Extension (MV3), AWS Amplify, AWS Bedrock (Claude), OpenAI
Similar Case Studies
More work from similar industries, or other recent stories from our library.
One authenticated dashboard that pulls every marketing channel into a single monthly view
How we built our own marketing dashboard, one authenticated app that pulls every channel into a single monthly view.
Structured Copilot enablement for ~2,500 service staff across two license tiers
How we brought structured Copilot training to roughly 2,500 service staff across two license tiers, then layered discovery, a power-user cohort, and train-the-trainer to make it stick.
Secure AI contract and spec review that gives 200+ hours back to estimating
How we built a dedicated platform that turns 50-page GC subcontracts into prioritized redlines in minutes, on the company's own secure infrastructure.
Tell us your biggest problem.
We'll show you the ROI.
Drop the problem in the box. A Wyecliff partner replies inside one business day with two ideas you can ship in 30 days. No pitch deck, no sales call required.