Published 06 Mar 2026  ·  Depesh Vyas

What AI Can and Can't Do For Your Operations (An Honest Breakdown)

There's a version of this question that comes up in almost every conversation I have with founders right now. Sometimes it's asked directly. Sometimes it's implied. But it's always there: can't I just use AI for this?

It's a fair question. AI tools have gotten genuinely good at a lot of things. And if you're running a $15K–$40K MRR service business, you're probably already using some combination of ChatGPT, Claude, or something similar for writing, summarising, or answering questions. So it's reasonable to wonder whether the same tools can handle the operational work that's currently consuming most of your week.

The honest answer is: some of it, yes. Most of it, no. And the part it can't do is almost always the part that's actually keeping you stuck.

Here's a clear breakdown.

What AI Is Genuinely Good At in Operations

Let's start with where AI actually earns its place, because dismissing it entirely would be wrong — and it would also tell you I haven't been paying attention.

Drafting documents and templates. SOPs, onboarding checklists, job descriptions, meeting agendas, email templates — AI can produce a working first draft of almost any operational document in a few minutes. If you know what you want the document to say and you're good at editing, AI cuts the time to produce it by 70–80%. That's real.

Summarising and processing information. Meeting notes, long email threads, client briefs, feedback — AI is very good at extracting the relevant points and producing a clean summary. For founders who are drowning in information, this is genuinely useful as a daily tool.

Answering process questions. "How should I structure a client onboarding flow?" "What should a weekly standup agenda include?" "What KPIs should a B2B agency be tracking?" AI will give you a reasonable answer to all of these. Not always perfectly calibrated to your business, but a solid starting point.

Basic automation and workflow logic. With tools like Zapier, Make, or even just native AI features in project management tools, a lot of repetitive operational work — task creation, status updates, follow-up reminders — can be automated. This is real time savings, and it's accessible without technical expertise.

So yes. AI is a legitimate productivity tool for the document layer of operations. If you're not using it for these things, you should be. It's not a question of whether to use it — it's a question of understanding what it can and can't do before you assume it replaces something more fundamental.

Where AI Stops Being Useful

Here's where the gap between what AI can produce and what your business actually needs becomes very clear.

AI can draft an SOP. It can't make anyone follow it.

This is one of the most consistent problems in service businesses at this stage. The SOP exists. It's in a Google Drive folder. It's reasonably well-written. Nobody uses it. The team still asks the founder every time something comes up. Nothing has changed.

The reason SOPs fail in small businesses is almost never the quality of the document. It's that there's no accountability system built around it — no peer review process, no definition of done, no consequence for not following it, no forum where deviations get caught and corrected. Building that accountability system requires someone with authority in the business to build and enforce it. That's an operator, not a document.

AI can tell you what a good operating cadence looks like. It can't run one.

If you ask an AI tool to design a weekly operating rhythm for a $20K MRR service business, it will give you something reasonable. Daily standups, Monday planning, Friday retrospective, weekly KPI review. Good structure. But the reason most founders don't have this in place isn't that they didn't know what it should look like — it's that nobody has built it, nobody is holding it, and every time it starts to take shape, the week gets busy and it collapses.

An operating cadence only works if it's non-negotiable. Making it non-negotiable requires someone in the business with the authority and presence to enforce it. That is not an AI problem. That is a leadership and accountability problem.

AI can generate a hiring framework. It can't evaluate whether the person in front of you is the right hire.

Job scorecard, interview questions, evaluation rubric — AI can build all of that. What it can't do is sit in the interview and notice that the candidate's answers are technically correct but something about how they talk about their last team is a flag. It can't tell you that the person you're about to hire will likely recreate the same dynamic you're trying to move away from. Hiring judgment at the $10K–$40K MRR stage, where one wrong hire can set you back three months, is pattern recognition built from experience. AI doesn't have that pattern recognition for your specific business in your specific context.

AI cannot tell you that you are the problem.

This is the biggest one. In a large number of service businesses at this stage, the founder is the primary operational constraint. Not the team, not the systems, not the market. The founder. Because every decision routes back to them, because they override the systems they build, because they can't actually let go of the quality gate even when they say they want to.

An AI tool will never tell you this. It will give you frameworks for delegation, answer your questions about how to build a better team, and generate a plan for stepping back from operations. But it will not look at how you're actually operating and tell you directly: the way you're running this business is the reason it isn't growing.

That conversation requires someone who has watched enough founder-led businesses to recognise the pattern, has enough standing in your business to say it credibly, and has enough skin in the game to be motivated to say it at all. That is not a tool. That is a person.

The Deeper Issue With "I'll Use AI for This"

There's something worth naming here, because it comes up often enough that it's worth being direct about.

When founders say they'll use AI to handle their operations, what they usually mean is they'll use AI to produce the documents, templates, and frameworks that represent operational infrastructure. And those things are real — they have value. But they are not the same as operational infrastructure that actually functions.

The difference is execution and accountability. A business that has a well-designed SOP that nobody follows is not operationally stronger than a business with no SOP. A business that has a weekly standup that collapses every third week because something more urgent comes up hasn't actually built an operating cadence. A business where the founder has "delegated" quality review but still checks every piece of work before it goes out hasn't removed the bottleneck.

AI can help you produce the artefacts of operational infrastructure. It cannot produce the conditions — the accountability, the authority, the enforcement, the consistent presence — that make those artefacts actually function.

This matters because founders who use AI to build a set of operational documents sometimes conclude they've fixed their operations when they haven't. The documents exist. The behaviour hasn't changed. Six months later, the business is in the same place — but now there's a Notion folder with twenty SOPs in it.

How to Think About AI and Operations Together

The right mental model is this: AI compresses the time it takes to produce operational documents and templates. It does not compress the time it takes to build a functioning operational system.

Building a functioning operational system requires diagnosing what's actually broken — which means looking at how decisions really get made, not how you think they get made. It requires building accountability structures with real teeth. It requires someone who will push back when the founder reinstates the behaviour the system was designed to eliminate. And it requires consistent presence over weeks and months, not a single session of document generation.

AI is a useful tool inside that process. It saves time at the document layer. But it is not a substitute for the process itself.

If your business is at $15K–$40K MRR and the operational ceiling is real — if you're working too many hours, the team is escalating everything to you, and scaling feels unsafe — the question isn't whether to use AI. The question is whether you have someone in your business who owns the operational function and is accountable for the outcomes. That is the gap AI doesn't fill.

---

Depesh Vyas is the founder of VBOG (Vyas Business Operations Group). He works with B2B service business founders at $15K–$40K MRR who are stuck in execution and want to build the operational infrastructure to grow without it all depending on them. The starting point is a $500 Operations Audit — a 7-day diagnostic that tells you exactly where your operations stand and what needs to be built first.