How Human-I-T saved 1.5 hours a day on data validation with AI Teammates

At Human-I-T, our mission is straightforward: get technology into the hands of people who need it. Every device that moves through one of our facilities has to be logged correctly with the right specs, the right format, and the right destination. When that data is wrong, devices get misrouted and reports break down. As Head of IT, it’s my job to make sure that never happens.

For a long time, that meant hours of manual review every day. Now, I lean on an AI Teammate I built in Asana to check and fix every device record before errors have a chance to spread.

When manual data entry can't keep pace with volume

Human-I-T processes hundreds of devices per day. Across a 28-person operations team, we manually enter technical specs like RAM, CPU, storage, form factor across hundreds of individual tasks. Small inconsistencies are inevitable at that volume. One person might enter a form factor one way, but the next would do it differently. A storage size could be formatted correctly in one task, wrong in the next.

My teammates and I were spending up to two hours every day on manual review, and we still weren't catching everything. The problem wasn’t the effort; it was that no human process could scale to match the volume. It was literally impossible.

Building an AI Teammate that validates device data before devices leave the warehouse

To both reduce errors and take back time, I built an AI Teammate in Asana that operates as a validation layer. I built it entirely using natural language instructions, no code required, and it sits between data entry and the rest of the pipeline. Here’s how it works.

Step 1: AI reviews every field on every device task

The moment a device enters one of four pipelines, the AI Teammate starts validating that data and formatting is entered correctly. It reads every asset field and checks each value against expected formats and known constraints. This happens automatically, for every device, every time.

Step 2: AI auto-corrects and normalizes inconsistent inputs

Rather than flagging errors for me to review and fix, the AI corrects them on the spot. Fields that used to be subject to human interpretation—like form factor—get filled in correctly because the AI can cross-reference the actual model and serial number to determine the right value. Knowing the model and serial number, it formats everything the way it's supposed to be with zero errors.

Step 3: Clean data flows downstream without human intervention

Once validated, the task is fully formatted and ready to move through the rest of the pipeline to a recipient or to Human-I-T's online store without anyone needing to review it manually. The AI runs 14-plus hours a day, continuously processing all the devices autonomously.

From two hours to 30 minutes and far fewer errors

Manual review dropped from two hours a day to 30 minutes but the time savings aren't even the real story. We eliminated the class of errors that was breaking our downstream reporting. Now, I can assess every device in our organization with confidence.

With that confidence, my role has changed. Time that I used to spend checking and fixing data now goes toward decisions that actually move the mission forward, things like finding better vendors, improving the tools the team relies on, and figuring out where AI can help next.

It's allowing our staff to be more proactive and less reactive. It's allowing me to start looking at more impactful areas of the business.

Extending AI to every high-volume workflow

This is just the first layer of what Human-I-T is building. Next, we’re expanding the same approach to other high-volume, error-prone workflows across the organization, moving closer to an AI-first operations model. My one piece of advice for anyone thinking about getting started: don't begin with the AI. Begin with the problem. Find the process that is breaking your data, your reporting, or your team's capacity. Then design the AI to solve that specific problem. That's where the real impact lives.