Architecture Firm Software: Stop Finding Fee Overruns 30 Days Late

Why project management software makes or breaks an architecture firm

The 70 person, three studio firms we work with carry somewhere between 40 and 90 live projects at once, and every one of them is a small business with its own fee, its own phase plan, its own consultant team, and its own way of quietly going upside down. The tools holding this together are never one tool. Timesheets sit in Deltek Vantagepoint, BQE Core, or Ajera. The drawing set lives in Revit and Bluebeam Revu, published to Autodesk Construction Cloud or Newforma, or shoved into the contractor's Procore. Phase plans and staffing live in a Smartsheet grid or a Monograph board. And the number that actually decides whether the year is profitable, fee spent by phase against real percent complete, lives in a spreadsheet on the studio director's laptop that gets rebuilt every Monday morning.

The pattern is always the same. It is Thursday. The controller runs last week's work in progress report. The K-12 project that was supposed to be 60 percent through construction documents at $1.4M of a $2.4M fee is actually at 74 percent of fee spent and roughly 55 percent complete by the project architect's honest read. That gap opened four weeks ago. Nobody caught it because timesheets closed Monday, the report ran Wednesday, and the project architect never opens the report anyway. By the time the principal hears about it, the firm has already given away close to $90k of unbillable labor, and the only remaining moves are to eat it or to open an additional services conversation with a client who thinks everything is fine.

That lag is the whole problem. Accounting systems are financially accurate and operationally blind: they report what happened after it stopped being fixable. Then add construction administration, where RFIs land in the contractor's Procore, get answered by a project architect over email, get logged in a spreadsheet if someone remembers, and burn 20 to 40 hours a month against a phase that was priced at 12. Nobody bills the overage because nobody can prove it happened.

Problem: fee burn reaches the controller 30 days late and never reaches the person spending it

The project architect who commits the money has no live view of it. She staffs two job captains onto CD coordination for a week because the structural drawings changed, and that decision costs about $14k. She finds out it was the wrong call five weeks later, in a principal meeting, from a PDF.

Vantagepoint and Ajera cannot fix this because their unit of truth is the accounting period, not the decision. Their percent complete field is typed in by a human once a month, usually optimistically, usually the day before invoicing. Monograph gets closer with visual planning, but its forecast is only as honest as the plan someone last touched, and it does not know that the roof detail set went into its third redesign.

What a custom build does: it makes burn a daily, per phase, per person number that shows up where work happens. Time entries stream from Vantagepoint nightly (or get captured directly and pushed back), map to a phase and task in your own work breakdown, and drive an estimate at completion that updates every morning. The project architect opens her project and sees CD at 71 percent of fee against 58 percent complete, with a projected overrun of $76k at the current run rate and the two weeks where the divergence started. Thresholds fire: at 15 percent variance the studio director gets it, at 25 percent the principal does, with the phase, the fee, the delta, and the last four weeks of hours attached. This is the single highest return feature we ship in this category, and it is usually live inside the first eight weeks.

The useful AI here is forecasting, not chat. Trained on your own closed projects, a model that knows project type, delivery method, square footage, client type, and consultant mix will flag at week three of CD that this project is tracking like the three previous school district projects that all landed 20 percent over, long before the linear burn line looks scary.

Problem: the drawing set and the project record live in different universes

The sheet index in Revit says one thing. The transmittal log says another. The contractor is building from a set issued nine days ago that included two sheets nobody logged. When a dispute lands, the contract administrator spends a day and a half reconstructing what was issued, to whom, and when, from Bluebeam sessions, email, and an ACC folder.

Newforma and ACC each hold part of this, and neither holds the fee. Procore holds the contractor's version of the record, which is exactly the version you do not want as your only copy when an ASI turns into a claim. None of them tie an issued sheet to the hours it cost or the scope it changed.

A custom build makes the issue record the spine. Every issuance, DD set, permit set, bid set, ASI, bulletin, gets a record with the sheet list pulled from the Revit sheet index by the ACC or Revit API, the recipients, the date, the reason, and a hash of the PDF stored in your own bucket. That record links to the RFI or change that caused it and to the hours logged against it. When the contractor claims they never got sheet A5.02 revision 3, you answer in 40 seconds, not a day and a half. Document extraction pays for itself here: parsing the issued PDF set to auto build the sheet list and diff it against the previous issue catches the two sheets that quietly changed and nobody flagged.

Problem: CA eats principals alive and nobody bills for it

Construction administration is where the firm's most expensive people do the least tracked work. An RFI arrives in Procore at 4:50pm. The project architect reads it on her phone, thinks about it overnight, calls the structural engineer, drafts a response, and closes it in Procore two days later. Three hours of a $210 rate, unattributed, unbilled, invisible.

Procore is built for the contractor's workflow and their clock, not your fee. It will not tell you that your CA phase, priced at 8 percent of fee, is running at 14 percent across six active projects, or that one particular general contractor generates triple the RFI volume of anyone else you work with.

The build: pull RFIs and submittals from the Procore or ACC API into your own register the moment they open, route them by discipline to the right reviewer, and start a clock. Every response gets time attached, either logged or estimated from the review session. The register rolls up to a CA dashboard: RFIs per week per project, average response time against the contract requirement, and hours consumed against the CA fee. AI does two jobs here that hold up in practice. First, after hours triage: an RFI landing at 4:50pm gets classified, matched against the spec section and past answers on similar projects, and a draft response with citations waits for the architect at 8am. She edits and stamps rather than starting cold, which we see cut RFI handling time roughly in half. Second, follow up: automatic chasing of overdue consultant submittal reviews, which is otherwise a job the job captain does badly.

Problem: percent complete is a guess, so billing and staffing are guesses

Percent complete drives your invoice, your revenue recognition, and your staffing plan. In most firms it is a number a project architect types in under pressure the day before billing. It is wrong in both directions, and both hurt: understate it and you finance the client's project, overstate it and you book revenue you will hand back.

No off the shelf tool can derive it, because none of them know what your deliverable is. Percent complete for a CD set is a function of sheets released, details resolved, consultant coordination cleared, and QA/QC comments closed, and that definition is yours, not the vendor's.

A custom build encodes your firm's deliverable model. Phases from your B101 agreements break into a task tree with weights. Sheets and details from the Revit model report their own status. QC comments close against tasks. Percent complete becomes a computed number the project architect adjusts with a reason, not invents from scratch. That flows into an AIA style G702 and G703 application, generated with the phase breakdown your institutional clients demand, and into the staffing plan: the resource board shows who is on what next month against forecast demand, so the studio director stops discovering on Friday that four job captains are all committed to Monday.

Cost and timeline for an architecture build

Across 2,000-plus projects, Digital Heroes has landed a consistent pattern. A focused first release, live burn and estimate at completion, phase tracking, a real RFI and submittal register, and a nightly Vantagepoint or Ajera sync, runs $60k to $130k and ships in 12 to 16 weeks. A full platform, adding proposals and fee build up, resource planning, consultant coordination, the issue and transmittal record, AIA billing generation, and client portals, runs $150k to $400k phased over 6 to 12 months. Phase it. Firms that try to build all of it before anyone uses any of it are the firms that end up with a $300k system and a spreadsheet still open on the studio director's laptop.

What pushes price up in this category specifically: Deltek Vantagepoint integration depth, because the reporting side is straightforward and writing back time and billing is not, and older on premise installs often need a middleware layer and an ODBC path rather than clean REST. Revit and ACC integration, because sheet and model data is genuinely awkward and the Forge/APS work is real engineering. Multi entity accounting if your offices are separate legal entities. AIA G702 and G703 generation with client specific formats, which is fiddly and always custom. Procore API access tiers, which are a negotiation with Procore, not a line in your budget. And migration of eight to twelve years of historical projects, which is the item most firms underestimate by a factor of two.

Build or buy: the honest line

Buy, honestly, if you are a single office under about 15 people. Monograph plus QuickBooks will run you fine, and a custom build at that size is vanity. Buy if your project mix is homogeneous, small, and short, because the spreadsheet is not really lying to you at that volume.

Build when these signals show up, and they show up together. Someone is paid, formally or not, to rebuild reports every week. You have more than 40 live projects and more than one office. You have written off more than $150k in a year on overruns you found late. Your project architects have stopped opening the accounting reports entirely. You are paying for Vantagepoint, Newforma, a Procore seat, Smartsheet, and Bluebeam, and still cannot answer "which five projects are bleeding right now" in under an hour.

The position worth holding: do not rebuild the general ledger. Vantagepoint and Ajera are perfectly good at accounting and dreadful at operations. Keep them as the system of record for money, and build the operating layer on top, the layer that knows what a phase, a sheet, an RFI, and a consultant are. In our experience, firms that try to replace their accounting system fail, and the ones that build the layer above it get the payback inside a year.

How to choose a developer for architecture firm software

Ask them to model your project on a whiteboard before any contract. If they cannot draw project, phase, task, fee allocation, consultant agreement, issuance, RFI, and time entry, and explain how percent complete and estimate at completion are computed across them, they will learn on your budget. A developer who thinks a phase is a Kanban column will build you Asana with your logo on it.

Make them prove the integration, not describe it. Ask for a working read from your actual Vantagepoint or Ajera instance in the first two weeks, before the full scope is signed. Ask specifically how they handle the Revit sheet index and whether they have shipped against the Autodesk APS and Procore APIs. Integration is 40 to 60 percent of the effort on the builds we have done in this category, and it is where honest estimates separate from optimistic ones.

Check the compliance and records side, because your institutional and public sector clients will. Contract document retention terms, audit trails on every issuance and approval, SSO through your Microsoft 365 tenant, and clarity on where the data physically sits. If you carry professional liability with defense obligations, the issuance record is evidence, and it needs to behave like it.

And settle ownership before kickoff, in writing: you own the source, the repositories, the infrastructure accounts, and the data, with a documented handover from day one. Ask directly what happens if you stop working with them in month seven. If the answer is vague, that is the answer.