I’m Calling This One Gratitude, As I Head Into a Few Days Off
Hi all 👋
Heading into a few days off, gratitude feels like the right theme for me this week.
I'm grateful for what we're building, for the partners helping to make it happen, for the support we get from our team in Australia, and, most of all, for the clients who've trusted us to do great work for them. It genuinely feels like a new chapter.
What's Happening in AI
A Teradata study is making the rounds saying the real blocker for agentic AI isn't the model; it's the data.
Aligns with what I keep seeing in the field. On a recent AP automation project, the first issue wasn't AI; it was that the same cost centers were named differently across two systems, so nobody could say for certain which invoices belonged where. We had to clean that up and get the systems agreeing with each other before any AI agent could safely make a call on its own.
That's the pattern almost every time, not the exception. The study backs it up too: 77% of executives say 20% or less of their enterprise data is described well enough for an agent to use.
Most companies think they have an AI problem. What they usually have is years of disconnected systems nobody ever cleaned up.
So if you're evaluating agentic AI right now, skip the "which model" question for a minute and ask a harder one: could you actually describe your own data well enough for an agent to trust it? Most companies I talk to still can't answer that.
This Week's Case Study
One university client was processing roughly 500 international student applications a day, each taking an assessor 15 to 30 minutes in a manual Word document with no real audit trail. Fine at low volume; at 500 a day, it was the bottleneck holding everything else up.
We automated the assessment and wired it into their student application system. Processing time dropped by about 65%, and every decision now carries a full audit trail. It's easy to blame people when compliance breaks down, but usually the process was never built to handle this much volume.
→ Read the full case study.
A Useful Resource
One question I get quite often is: "How do I know if a process is actually worth automating?"
That's why we built the Administrative Tax Calculator. It shows you how much you're paying premium salaries to do low-value work, chasing updates, manual entry, bridging gaps between systems. Before you hire another person, see how much you're already losing to broken processes.
That's it for this week.
— Robbie
Get the latest tech news in your inbox
Hi all 👋
Heading into a few days off, gratitude feels like the right theme for me this week.
I'm grateful for what we're building, for the partners helping to make it happen, for the support we get from our team in Australia, and, most of all, for the clients who've trusted us to do great work for them. It genuinely feels like a new chapter.
What's Happening in AI
A Teradata study is making the rounds saying the real blocker for agentic AI isn't the model; it's the data.
Aligns with what I keep seeing in the field. On a recent AP automation project, the first issue wasn't AI; it was that the same cost centers were named differently across two systems, so nobody could say for certain which invoices belonged where. We had to clean that up and get the systems agreeing with each other before any AI agent could safely make a call on its own.
That's the pattern almost every time, not the exception. The study backs it up too: 77% of executives say 20% or less of their enterprise data is described well enough for an agent to use.
Most companies think they have an AI problem. What they usually have is years of disconnected systems nobody ever cleaned up.
So if you're evaluating agentic AI right now, skip the "which model" question for a minute and ask a harder one: could you actually describe your own data well enough for an agent to trust it? Most companies I talk to still can't answer that.
This Week's Case Study
One university client was processing roughly 500 international student applications a day, each taking an assessor 15 to 30 minutes in a manual Word document with no real audit trail. Fine at low volume; at 500 a day, it was the bottleneck holding everything else up.
We automated the assessment and wired it into their student application system. Processing time dropped by about 65%, and every decision now carries a full audit trail. It's easy to blame people when compliance breaks down, but usually the process was never built to handle this much volume.
→ Read the full case study.
A Useful Resource
One question I get quite often is: "How do I know if a process is actually worth automating?"
That's why we built the Administrative Tax Calculator. It shows you how much you're paying premium salaries to do low-value work, chasing updates, manual entry, bridging gaps between systems. Before you hire another person, see how much you're already losing to broken processes.
That's it for this week.
— Robbie
