A field tech logs a site visit in a mobile data collection app. Later that day, the same visit gets typed into a timesheet. Then into a project tracker, so the PM can see hours against budget. Then a third time, by someone in accounting, onto an invoice line. Four systems, one site visit, four separate typing jobs, and four separate chances to get the hours or the task code wrong.
That happens because the firm’s technical stack and its business stack were bought at different times, from different vendors, and nobody ever wired them together. Training won’t fix it, and neither will discipline.
A pattern, not an exception
We saw this in three-quarters of the 48 environmental consulting firms we reviewed across the western US, and mostly as the default way data moves through the firm rather than an occasional workaround. GIS platforms, modeling software, and field data collection apps on one side. Time capture and billing on the other. The two sides share neither a database nor an API, and in most cases not even a common project ID, so a person becomes the connection between them.
The failure mode shows up even when a firm buys its way to bigger systems. The CTO of one environmental consultancy we interviewed had just been through an ERP migration meant to solve exactly this, and described the result: “This is two weeks later and I still can’t get any reports on company revenue. It’s all a big box.” A larger system that nobody can see into is the same silo with a higher licence fee.
Nor is the underlying condition unique to environmental consulting. A March 2026 survey of 1,389 US social-impact professionals by the Blackbaud Institute with Edge Research (a different sector, so treat it as corroboration rather than industry data) found fewer than 20 percent of AI-Emerging organizations would rate their own data health as excellent, roughly doubling among the most AI-mature. Those mature organizations averaged 6 dedicated data staff against roughly one elsewhere. Small consultancies don’t have one data staffer, let alone six, which is exactly why the person carrying data between systems is usually a project manager or a principal doing it on top of a billable job.
Fix the biggest retype, not the whole stack
None of this is a case for ripping out the technical stack and starting over. GIS shows up in the large majority of technical job postings we looked at for a reason: it’s the right tool, and firms that run it well should keep running it. The tools are fine. The gap sits between them and the systems that track hours and send invoices, and today a person has to stand in it and carry the data across by hand.
That’s also why “buy more software” is usually the wrong first move, and it’s the mistake I see most often. A firm notices the retyping pain, goes shopping for a fifth system that promises to unify everything, and six months later has five systems that don’t talk to each other instead of four. The ERP story above is the expensive version of the same move. Make the systems already in place pass data to each other automatically instead, even if that only means one well-built connection at the worst bottleneck rather than a full rebuild.
The place to start is wherever the same number gets typed the most times before it reaches an invoice. For most of the firms we looked at, that’s hours: a technician’s time gets captured once in the field, then again in a timesheet, then a third time in a project tracker, and none of those three entries update each other. Fixing that one link usually returns more time than any other single change, because it’s the link every project touches every week.
What one well-built connection actually involves
Three routes, in the order worth checking them. First, the native integration. Many field data and time-tracking tools already ship connectors for the common project management and accounting systems; firms often own an integration they’ve never switched on. Check both vendors’ integration pages before assuming you need to build anything.
Second, middleware. Where no native connector exists, the workflow-automation platforms most firms already recognize can watch one system and write to another. This suits one-directional flows, like field hours landing in the timesheet, and it’s a configuration job measured in days, done by someone comfortable with the tools rather than a software engineer.
Third, a small custom job: a script against the two systems’ APIs, worth it when the middleware route can’t express the logic (say, mapping your field app’s task codes onto your billing structure). Still a small, bounded piece of work when it’s scoped to one link. The trap in every route is scoping it to “integrate everything,” which turns a well-built connection into the failed ERP project. Whichever route fits, hold it to the same three requirements: data flows in one defined direction, one system is agreed as the source of truth for that number, and errors land in a log somebody actually checks.
The twenty-minute exercise that finds your worst link
Before talking to anyone about any of this, run the source-of-truth test. Write down four numbers your firm depends on: hours worked, project status, budget spent, invoice status. For each one, name the single system where the authoritative version lives. Where you can’t, or where two people name two different systems, you’ve found the silo that matters, and almost always the retype that goes with it.
The firms that fix this fastest are the ones that can pass that test in one sentence per number. Most firms we’ve worked with can’t, and that ambiguity keeps the retyping going long after everyone has noticed the pain. I run into this from the operations side, not the GIS side, which is where GaiaOps sits: the fix is rarely a new tool, and usually a decision about authority between the tools already paid for.
The Digital Operations Maturity Rubric will show you in about ten minutes where your firm sits on this; it’s a self-score built from the same research. The data behind the numbers above is in What 48 Environmental Firms Taught Us About Operations, and how GaiaOps prices the fix is on the pricing page.
Or start with the retype count itself. A diagnostic call maps where your systems stop talking to each other in thirty minutes, using your numbers rather than the survey’s.