On July 17th, OpenAI released its much-anticipated ChatGPT Agent. Watching the announcement, one thing became clear: the gap between demos and enterprise reality is as wide as ever. The demonstrations, while impressive for consumers, don’t reflect how most professionals in tax and accounting will actually use these tools.
Nonetheless, agents are here, and the same question that plagued people this time 2 years ago gets repeated: “How do I put this to use”? Two years ago with the GPT-4 chatbot, it was “make me a recipe” and “design my workout”, today it’s “buy a suit, shoes, a wedding gift, then book a flight and hotel” but professionals aren’t booking trips and shopping for wedding attire multiple times a day.
I’ve now been using agentic tools for a few months: coding with Windsurf.ai, Manus for slides, and my thought partner “Kyle-EM” (a sort of digital thought-twin imbued with my professional experience) runs on a highly agentic platform. Well-orchestrated* agents are incredible to work with.
*poorly orchestrated agents are a disaster.
What Agents Can Really Do—Right Now
Let’s be clear about what’s possible today: Agents can compile information, access structured data, automate web searches, fill out straightforward forms, create draft spreadsheets, and even generate simple slide decks. This cuts down on repetitive, manual chores and reduces the “grunt work” burden.
But there are sharp limits.
- Today’s agents may create a spreadsheet, but they simply can’t build out full tax workpapers: those complex, linked, multi-sheet files with formulas, documentation links, and dynamic referencing.
- They aren’t ready to manage nuanced reconciliations, exception handling, or the detailed flow that tax teams rely on within workpapers and reporting packages.
- They’re not going to be let loose to log in to your ERP to start pulling reports and gathering source data out of systems of record (yet).
2027 Is Coming—So What Should We Do Until Then?
Top AI CEOs and many other credible commentators predict 2027 will be the year when properly secure and capable agents are rolled out in waves across enterprises.
With the advancing capabilities of agentic AI, it’s natural to wonder:
When agents are able to handle end-to-end processes, will we simply have to “tough it out” until that point arrives, waiting for the technology to mature enough to take over the bulk of our work?
This is the critical question for every tax and finance leader today:
“Do we keep working as we always have, crossing our fingers for future disruption?”
Or do we invest now in accelerating our digital transformation—so these agents have a solid foundation when they arrive?
Start by reading this 3 minute post by my friends at Innovation Algebra.
Automating Chaos: Is It Possible?
It’s tempting to believe that agents will eventually be able to handle all the messiness for us. The reality is more nuanced. Strategic automation using traditional approaches, supported by today’s increasingly capable AI tools, can clean up a surprising amount of data sprawl and workflow chaos. With smart prompting, business-led process mapping, and AI-driven code (e.g. macros) and documentation generation, major pain points can start to be smoothed out right now.
But this outcome is not automatic. Automated chaos is still chaos, unless it’s guided by expertise and proactive investment in process improvement.
Accelerate Digital Transformation Starting Now
You don’t need to scramble together an AI automation program. Let the “flatness” of this moment do its work.
That said, if you want tomorrow’s AI agents to function at an expert level inside your tax department, you need to build the conditions for their success. That means moving digital transformation work from the back burner to the top of your agenda.
- Map and document your key workflows. Let AI tools help capture steps, surface inefficiency, and standardize documentation.
- Tackle data hygiene. Use automation such as macro scripts, AI cleaning routines, validation frameworks, alongside human review to elevate the quality and consistency of your core data assets.
- Modularize your processes. Break up complex end-to-end work into defined segments and automate what you can now. Every script or template you create is a “starter kit” for future AI integration.
- Upskill your team with project-based learning using AI-assisted automation. AI isn’t just for coders. Let practitioners ask questions, generate code, build flowcharts, and validate steps using AI assistants.
- Don’t just automate—reimagine. Question the value and design of every process you’re tempted to automate. Use transformation as an opportunity for simplification and systemic improvement, not just speed. Don’t automate garbage.
Prepare Today for Agentic Gains Tomorrow
The pace of progress is real, and the allure of powerful agents is strong. But there is no leapfrog without a runway. The foundational work you do today—standardizing, mapping, and automating—won’t just ease your daily pain. It will create the conditions for agents to propel your function forward when they’re finally ready for production-grade enterprise work.
Start now, so that when the future arrives, you, and your tax team, are out in front.
