AI Efficiency for SMBs: Why AI and Automation Work Better Together
As consumption-based AI licensing becomes the norm, inefficient use of AI starts to cost real money. The value comes from knowing where AI belongs in a process, and where automation is still faster, cheaper and more reliable.
I think we're hitting a bit of a turning point with AI at the moment.
For the last couple of years, it's felt like you could just throw anything at it, tweak it, try it again, tweak it again, and eventually you'd get something relatively decent out the other end.
But that way of working is starting to get very expensive.
Every time you "have another go", you're effectively paying for it. And as consumption-based AI licensing, through credits, tokens or usage, becomes the new normal, inefficient use of AI could start costing SMBs heavily.
AI Efficiency Is Becoming More Important
A simple way I've been thinking about it is a bit like fuel.
Years ago, when fuel was cheap, you didn't really worry too much about efficiency. As long as the car went, that was good enough. But as fuel prices have risen, efficiency has moved front of mind.
AI is heading in exactly the same direction.
The infrastructure behind AI is expensive, and those costs are starting to show up in how tools are priced and limited. So businesses can't necessarily treat AI as an unlimited resource anymore.
For SMBs in particular, that means thinking more carefully about AI efficiency. Not just whether AI can perform a task, but whether it's the most efficient and cost-effective way to do it.
Where AI Works Well and Where It Doesn't
Where I see a lot of businesses facing challenges is trying to use AI for end-to-end tasks like document creation.
They'll take a meeting transcript, some emails, a few bits of input and simply say: "Create me the final document."
But you can actually break that process into two very different tasks.
AI: turning unstructured data into useful content
Interpretation, summarisation and structure, from information that would otherwise take significant manual effort.
Automation: producing a consistent output
Defined rules, on-brand templates and the same reliable result every single time.
Using AI to Turn Unstructured Data Into Useful Content
The first part is taking all that messy, unstructured information and turning it into something meaningful, and AI is brilliant at that.
It's very, very good at taking fuzzy data, understanding the context and distilling it down into something that reads well and makes sense.
This is where AI can add huge value to a business process: helping people summarise, structure, analyse and create a strong starting point from information that would otherwise take significant manual effort to work through. Tools like Microsoft 365 Copilot sit naturally in this part of the process, because they already have context on your emails, meetings and documents.
Using Automation to Create Consistent Outputs
The second part, the actual production of the document, is where things start to break down.
Most businesses don't want something creative or different every time. They want consistency. They want it to look right, feel on brand and be repeatable.
And ironically, that's the bit AI isn't particularly good at, or at least not reliably.
In testing, we've seen the content creation part take seconds, while the document generation can take minutes, and even then it's not always right.
That's where automation still absolutely comes into its own.
If you separate those two steps, using AI to create the content and then automation to put it into a structured template, you get something that's faster, cheaper and consistent. This is exactly the kind of work Power Automate and the wider Power Platform were built for.
We've seen automation complete that second part in a few seconds, compared with several minutes using AI, and do it reliably every single time.
Why AI and Automation Should Work Together
This is why I don't think businesses should be looking at AI and automation as competing technologies.
They're good at different things.
AI is particularly powerful when a process requires interpretation, summarisation or working with unstructured information. Automation is powerful when a process needs to follow defined rules and produce a consistent result every time.
Combine the two and you can create a much more efficient process. Use AI where intelligence and flexibility add value, then hand over to automation where speed, consistency and reliability matter more. That combination is the basis of our AI & Process Automation work with SMBs.
There's Also the Question of Perceived Value
There's also a slightly softer point here, which I think is becoming more important: people are starting to feel AI content.
Not necessarily in a technical sense, but you can tell when something feels a bit generic or a bit "off".
And if you're a service business charging for your experience, that matters. The perception of value can drop if it looks like a machine did all the work, even if behind the scenes you've used AI to enhance what you're doing.
Using AI effectively doesn't have to mean making AI visible in every part of the end result.
Sometimes the better use of AI is behind the scenes, helping your people work faster and make better use of information while keeping the final customer experience consistent and recognisably yours.
The Question Isn't "Can AI Do This?"
So for me, the shift is pretty simple, but quite important.
It's less about asking:
And more about asking:
For SMBs, in a lot of cases, it's going to be at the front, helping you think, summarise, structure and get to a good starting point.
Then you hand over to something more deterministic, like automation, to actually execute and deliver the final output.
Designing a process that way also means being clear about what data the AI is working from. We covered that groundwork separately in preparing your data for AI and automation, and the governance risks in AI agents and tomorrow's shadow IT.
As AI costs rise, I think we'll see a slowing of what has been a very experimental phase in the AI journey for small businesses. Where AI has been a bit of a free-for-all, we'll start to see something much more intentional: processes that are designed, structured and ultimately more efficient.
Don't Forget About Automation
So, in summary: don't forget about automation.
AI is brilliant, but it's not always the right tool for the job. In a lot of cases, automation is still faster, cheaper and more reliable.
The real value comes from combining the two properly: AI to do the thinking, automation to do the doing.
If you get that balance right, you get the best of both worlds without blowing your budget on inefficient AI.
Want help deciding where AI belongs in your processes?
Book a discovery call to talk through the processes you want to improve, or get your TrulyFIT Score for an instant view of where technology is holding you back.
