If your site's still doing manual data entry in 2026, you might sometimes ask yourself why "computers" haven't come up with a modern solution yet. The truth is many software vendors have tried their version addressing the problems with manual data entry (it's messy and time-consuming, and the smallest errors easily snowball into dreadful series of queries), but none of those approaches have properly checked all of the boxes, especially for the sites actually using the tool. The proportion of sites still doing manual data entry today is not surprising, but automation has come a long way. So now might be the best time to ask yourself if manual data entry is still a good idea compared to software-assisted smart data entry specially made for clinical research sites.

The work we've quietly accepted

Picture a coordinator on a busy screening visit. At the bare minimum they have two browser tabs open — the eSource and the EDC — but it's not uncommon for lab portals, PDFs, and other sources and systems to be referenced . For the next hour, the job is transcription: read a value here, type it there, match the unit, confirm the date, then do it again, field after field, visit after visit. It's exacting work, and it's really not the best use of a trained coordinator's time.

This isn't any particular party’s failure, it's just the nature of the task, carried for years by capable people who never had quite the right tool for it. But it's worth naming the real cost, because almost none of it shows up as a line item. Every hand-typed character is a chance for a small slip, and one shifted digit or wrong unit can become a query that pulls a coordinator off their real work days later to research, fix, and re-verify. The real cost of manual data entry goes far beyond just the hours coordinators spend typing in data. Every manually typed field is an opportunity for your site to get dinged on data quality that impacts your ability to get the next study. You pay in coordinator hours, in evenings spent catching up on data entry after the last patient leaves, and in the slow burnout that sends your most experienced people looking for a clinical research site that respects their time.

So what would it actually take to beat doing it by hand?

Plenty of vendors have promised to fix data entry, and your coordinators have probably been handed a few of their solutions. The reason so many research sites still work by hand is that most of those tools cleared one bar and missed the rest. To genuinely beat manual entry, a tool has to win on every axis that matters, at the same time.

It has to be dramatically faster, moving source into the EDC without the tab-switching marathon rather than trimming a second off each keystroke. It has to be more accurate, making typos and other human errors much rarer, instead of just failing faster. It has to prevent queries at the moment of entry, checking ranges, units, dates, and required fields as the work happens, and flagging the quieter risks: a source that doesn't actually contain what the CRF is asking for; or a gap between what the source says, what the protocol requires, and what's about to land in the EDC. It has to read from any source and fill any EDC, because even though sites get to choose the source, sponsors decide on the EDC,  and your coordinators have to live with the result. And it has to do all of that while keeping a qualified person in control of what gets submitted. If you miss any one of these, you haven't replaced manual entry... you've just added another login on top of it.

A bicycle for the coordinator

Coordinators aren't going anywhere, and the rules make sure of it. Source data doesn't reach the EDC until an accountable person signs off on it, and that person has to be a person. Your CRCs and site staff are the unsung heroes of clinical research, and no tool is going to replace them. But a person can only realistically do a human volume of work, and asking them to do more transcription, faster, doesn't change that ceiling. It just wears them down.

There's an old Steve Jobs comparison I keep coming back to: on their own two feet, a human is a fairly unremarkable mover. But put that same human on a bicycle and they completely outperform the most efficient/fastest-moving species on the planet, "completely off the top of the charts", proving to be roughly twice as efficient as the next closest competitor. The bicycle doesn't replace the person or make their decisions for them; it multiplies what they can already do. Good automation is a bicycle for the coordinator, and ultimately for the site. The software does the mechanical part at machine scale: it studies what each field expects, fetches the source, and fills the entry into the CRF. The coordinator does the part only a trained person can do: confirm it against source and sign. Your team confirms; the software types and transcribes. That division of labor is what lets a coordinator clear a heavy screening visit in the time they used to spend on a single form.

The human stays in the loop, and on the record

This is the line between an assistant and an autopilot, and it's the one I care about most. A tool worth adopting fills the CRF and then hands off approval, staging every value for a trained person to review, approve, and sign before anything reaches the EDC. It never submits on its own. Accountability doesn't change: your research site is responsible for its data under 21 CFR 312.62 and the Form 1572, whether a coordinator types each character or confirms a fill.

The other half of staying on the record is showing your work. Every value the tool fills should link straight back to the exact source it came from, with a plain-language reason, so a coordinator, a data manager, a monitor, or an inspector can compare and verify it in seconds. That's the quiet difference between a tool that only saves time and one that also survives an audit. It's also the answer to the question we hear in almost every conversation: what will our sponsors think if they find out we're using assistance on their trial data? Their data teams want the same things you do, namely accurate data, a clean audit trail, fewer queries, and a faster path to database lock. A validated, human-reviewed, fully traceable tool delivers exactly that.

You can decide this one yourself

Here's what makes data entry different from the rest of your stack. A sponsor or CRO picks the EDC, validates it  over months, and hands it down to your research site as a requirement. A tool that simply helps your own coordinators do work they're already authorized to do doesn't have to travel that road. If it needs sponsor sign-off, an IT integration project, or a fresh validation cycle before anyone can touch it, you've wrapped a committee and a quarter around something that could take an afternoon.

Look for software you can turn on this week, for this study, without asking anyone. In practice, that means a browser extension your coordinators add to Chrome in a single click: no server to stand up, no integration to negotiate, and no change to how the source or the EDC behaves. The data-entry tool should ride along, inside the browser they already use, read from the systems they already have open, and fill the fields for them. Gleam is a proponent of the belief that the best clinical software fades into the background: it shows up exactly when it's useful and otherwise stays out of the way.

The 2026 verdict

So: manual entry, or automated assistance? For years the answer was "stick with what you can trust," because the tools couldn't clear every bar at once. That's no longer true. The clinical research sites already using Gleam today are reporting more than half of their data-entry time handed back, and on heavy screening visits, hours. The arithmetic your research site can do on the back of a napkin has flipped: what does an hour of coordinator time cost you, how many hours would you get back across your portfolio, and what is it worth to keep an experienced coordinator from burning out on after-hours transcription and leaving? More often than not, manual entry is now the more expensive option, not the safer one.

That's the lens we built Gleam around. Not one more set of empty fields to fill, or an entirely separate system to manage, but a smart data-entry assistant that rides along with your coordinators between source and EDC. It learns what each field expects, fetches from any source (including paper and PDF), fills any EDC, and traces every value back to where it came from, all with a qualified person confirming and signing before anything is submitted, and without asking the sponsor's permission first. Your team stays in control and on the record. Gleam just does the typing in between.