I Thought AI Would Replace Consultants. I Was Wrong.

Inside my experiment building Streamtech Advisors with AI agents instead of analysts.

Six months ago I assumed AI would replace consultants. Instead, it helped me rebuild one.

Over the past year something changed. The tools crossed a threshold where a single operator can do work that previously required a small research team. So I decided to test it.

For most of the past 20 years, consulting followed a predictable structure: a partner with experience, a few analysts, a research pipeline, and enough billable hours to support it all. It worked. Years ago I built PayStream Advisors, studying the systems behind accounts payable, payments, and financial operations. After the firm was acquired by Endava, I spent several years inside software companies trying to automate those same processes.

Which, in practice, usually means wrestling with ERP workflows, messy vendor data, and the endless edge cases that live between invoices, payments, and reconciliation. But the traditional structure always had friction. Research takes time. Analysis requires iteration. Ideas emerge slowly. And most clients ultimately want insight, not hours.

A Strange Signal in the Knowledge Economy

One of the data points that pushed me toward this experiment came from economist Gad Levanon, who writes the Substack Labor Matters. He’s been tracking something unusual: hiring in several of the core knowledge-economy sectors—finance, insurance, information, and professional services—has essentially flatlined since late 2022, something that historically almost never happens outside of recessions.

Those industries—consulting, accounting, finance, and tech—are built around work that involves analysis, writing, planning, and coding. In other words, exactly the kinds of tasks that, as Ethan Mollick has documented in One Useful Thing, AI systems are becoming extremely good at augmenting rather than replacing.

If that trend continues, the implications are significant. The knowledge economy may not be shrinking. But it may be learning to produce more without hiring proportionally more people. Which raises a question I couldn’t stop thinking about: What happens to advisory work when the marginal cost of research collapses?

That question eventually became Streamtech Advisors and The Velocity Stack.

The Experiment

I launched Streamtech Advisors as a new advisory practice focused on cash flow and financial operations systems. But the real experiment isn’t the consulting. The experiment is the operating model.

Instead of hiring analysts, I’ve been building what I jokingly call a small AI research lab. A set of tools and agents that help with:

• market mapping
• research synthesis
• data cleanup and enrichment
• draft analysis and writing
• visual frameworks

The stack currently includes Claude’s agentic capabilities, paired with orchestration tools like OpenClaw, along with a handful of smaller utilities. If that sounds complicated, the practical effect is simple. Questions that used to require days of research now produce structured insight in minutes. Not perfect insight. But highly usable starting points.

Building on Ethan Mollick’s framework for AI-augmented professionals, I think of it less as automation and more like having a team of tireless junior analysts who never sleep and never complain about revision requests.

What Actually Changed

The biggest shift isn’t speed. It’s iteration. Traditional research has a cost structure that discourages exploration. Every detour consumes billable time. Agentic tools flip that dynamic. You can explore five directions instead of one. Test hypotheses quickly. Refine frameworks before showing them to anyone.

It becomes possible to think more like a scientist running experiments than a consultant delivering a report. And that shift matters. Because many of the biggest problems in financial operations today—cash conversion cycles, invoice disputes, reconciliation complexity—aren’t solved by software alone. They require better thinking about systems.

Where This Shows Up in Practice

A simple example. When analyzing accounts receivable automation solutions for a client, a typical consulting workflow might involve: market research, vendor mapping, interviews, draft analysis, iteration.

That process could easily take weeks. With an agent-assisted workflow, the early stages collapse dramatically. Within a few hours you can:

• map the business process
• identify structural problems and time sinks
• develop workable solution frameworks

The human role shifts. Less time collecting data. More time interpreting what matters. While Alex Johnson tracks in Fintech Takes how fintech companies are trying to automate financial operations from the outside, I’m exploring how AI changes the consulting that helps companies implement these systems from the inside.

The Velocity Stack

To document what I’m learning, I started publishing this research series called The Velocity Stack. It focuses on the intersection of:

• financial operations and ERP systems
• payments and billing infrastructure
• automation, AI, and agentic workflows

The publication uses a three-layer framework to track where control of enterprise cash is shifting: the Legacy Layer (where cash gets trapped), the Challenger Layer (where the battle is happening), and the Control Layer (where the market is going).

But it’s also a running log of the experiment itself. What works. What breaks. What changes when a solo operator suddenly has access to capabilities that previously required a firm.

Some posts will explore industry dynamics. Others will document the tools and workflows behind the scenes. Because I suspect this shift is bigger than consulting. It may redefine how small, specialized advisory firms operate.

The Honest Truth

I don’t know if this model works long term. There are still obvious limits. Judgment still matters. Experience still matters. Relationships still matter.

But the early signal is clear. Advisors who treat AI as a threat will struggle. Advisors who treat it as infrastructure will build entirely different businesses.

If the industrial revolution created the modern firm, the AI era may create something else entirely: the one-person research lab.

Streamtech Advisors is my attempt to explore that frontier.

And if our worlds intersect—finance, payments, software, or AI—I’d enjoy reconnecting. I’m also curious what others are seeing: are AI tools actually changing your work yet?

The experiment is just getting started.

—
Henry Ijams
Streamtech Advisors