It is October. Which means a little anxiety can already be felt.
There is a sense of realisation that Q4 arrived before anyone was ready for it, that the year is nearly done and that somewhere in the diary there is a budget planning session for 2027 that you have been meaning to schedule since September.
We work with agency founders. And right now, the conversations we are having feel different to previous years. Not because budgeting has become harder, but because there is a new variable in the mix that does not behave like any other cost. AI has arrived in the P&L and nobody, including the agencies already using it well, may know how to account for it.
We have written this for founders who want guidance on how to manage the three most common costs and hopefully you will get into January without the massive panic of not knowing where everything is headed. This is highly iterative, so as always read this article as a one off and informational in nature.
Understanding the real picture from the data
The BenchPress 2026 report analysed over 800 agencies with revenues above £1 million. Profitability was improving across the sector, which is the good news. But the report also found that a significant number of agencies are still underpricing their work or simply do not have a clear picture of their true cost base. Which means the growth on paper is not always landing where it should.
That is the very commonly we recognise most from the conversations we have. The agency is busy. The team is stretched. Cash flows are tight,the founder looks at what is actually in the account at the end of the month and something does not quite add up. The margin is tighter than expected and they are not entirely sure why.
For most agencies in this bracket, three expenses are going to influence that more than anything in 2027.
Payroll | Marketing spend | AI tools
Payroll: the monthly number that does not move, even when everything else does
Payroll is always the biggest single cost in an agency. For businesses in the £1 million to £5 million range, let’s consider it somewhere between 40% to 65% of revenue.
So what is changing now?
Founders are more cautious about hiring than they were ever before. Not because work has dried up, but because they are not sure what AI means for how much capacity a team actually needs. The BenchPress data showed that only fifteen percent of agencies are using AI across the business in any meaningful, structured way. The rest are somewhere in between, with individuals picking up tools on their own, without any coordinated approach or training.
That gap is a problem when you are trying to plan a headcount for a whole year. The smarter question to address now might not be how many people do we need, but what does AI genuinely change about how we deliver and are we actually measuring it. Revenue per head is becoming the metric that separates agencies that understand their efficiency from those that are guessing. The ones doing the harder check on this now will be better placed to figure out where to spend what comes in 2027.
Marketing Spend: Paid ads and other areas
A lot of agencies in this revenue range are running paid ads for their own new business pipeline. And the ROI around that has changed a lot.
Average cost per click has risen significantly over the past two years. In competitive categories like marketing services, financial services and professional services, the increases have been steep.
Part of this is structural. More businesses are competing for the same search inventory and visibility. For example, Google’s own AI tools optimise bids upward. Reduced organic visibility pushing more of the competition into paid. The result is that the cost of acquiring a lead through paid ads has gone up significantly, without any corresponding improvement in the quality of those leads.
For an agency heading into 2027 budget planning, this is worth looking at honestly.
- What are your paid Ads actually returning right now?
- How does the cost per lead compare to what you were seeing in 2025 if you have this data.
- And if the numbers have changed, is increasing the budget the right answer, or is it worth redirecting some of that spend toward channels where the ROI are more predictable.
This is not an argument against paid Ads. For some agencies it remains a reliable source of enquiries. But for some, this can be quite run of the mill without an end in sight.
Additionally, efforts such as in person events/ networking and outbound activities cost a lot of money to organise once you think of it. Events are a category that is easy to undercount. There are the direct costs, venue, tickets, travel, accommodation, sponsorship where relevant. And then there is the time cost, which rarely makes it onto anyone’s marketing budget but is real and meaningful at founder or director level. Organising your own roundtable or client event costs considerably more than it looks once you count in venue, preparation time and follow-up.
Outbound efforts are worth accounting for properly too. The tools alone add up; it could be a LinkedIn Sales Navigator subscription, an email sequencing platform, a data source for contact lists, whatever CRM sits underneath it all. Add the time cost of prospecting, researching, writing and following up, outbound has a real price tag that rarely gets noticed in full. If nobody has added that up recently, the total tends to be higher than expected.
What all of these have in common is that they tend to live across different people, different budgets and different spreadsheets, which means the founder rarely has a clear total. Before you set a marketing number for 2027, it is worth knowing what you actually spent in 2026 across all of those categories and whether it was intentional and what was the return on it..
The AI budget that nobody knows how to build
And here is the big one and we will spend some time here. AI has become existential to most businesses on a day to day. However, many of us, though it;s hard to admit, are not sure where we are headed with this. In a recent podcast we listened to, this is considered well and there are lots of great tips about how to think about this.
This is the part of the conversation that feels most uncomfortable and it is probably the one most worth having. AI is already a cost in your business. Even if nobody has put it on a spreadsheet. Subscriptions to Claude, ChatGPT and a range of other tools are sitting across multiple team members, often on personal cards or bundled into a general tech budget or monthly allowance that does not show the individual lines. And as one of few pieces of information shared on this topic is on our Linkedin Post, the spend is happening whether anyone is tracking it or not.
Subscriptions are just the visible part. As agencies use these tools more, they start running into consumption based costs. Token spend. API usage. Cloud infrastructure that charges in the background of tools you thought you had a fixed price on. These costs can grow in ways that catch people off guard, particularly if there is no policy around how AI tools are used and by whom.
Research from Ramp found that around 31% of companies are already spending more than $10,000 a month on AI tools. And yet only twelve percent have any real structure in place for budgeting it. Most businesses are somewhere in the middle: using the tools, seeing some benefit, but without a clear way to track what it is costing versus what it is actually changing.
As was mentioned in the Equity Funcast (podcast episode): around 80% of AI pilots fail to deliver measurable results. Not because the tools are poor, but because the initiatives do not have clear ownership, a defined goal, or any mechanism to track whether it worked. Someone buys a tool. A few people use it occasionally. Nobody ever reviews whether it was worth it.
For a smaller agency, that is a real financial risk. You can carry a meaningful monthly AI spend with no visibility into what, if anything, it is returning.
Are you budgeting for an outcome ?
The agencies getting this to work well are not asking how much we should budget for AI. They are asking what do we expect AI to change and what would that change be worth to the business.
According to (reference) 20% reduction in the time spent on first drafts. Faster turnarounds on client reporting. Less manual time in financial reconciliation. These are outcomes that can justify a number. They give someone in the team a thing to own and a way to tell whether it is working or not.
Data quality is also worth thinking about before investing further. As was repeatedly spoken about in our client BBD Boom’s Elevate event of Sep 2026 (link Boom’s website), AI tools work off the information you give them. If your client data is inconsistent, your internal processes are not documented, or your financial data lives in someone’s head or in a spreadsheet that only one person understands, the output from any AI tool is going to reflect that. Cleaning up the foundations before scaling the tools is rarely the exciting part, but it is the part that tends to determine whether the investment pays off.
The adoption piece is perhaps the hardest. Technology is straightforward and ever growing. Getting a team to genuinely change how they work, to trust the output, to build new habits rather than just adding another tool to a long list of things nobody has time to learn properly, that takes someone with both the responsibility and the authority to actually drive it.
Where to start, practically
If budget planning is landing on your radar for the next few weeks, a few things are worth getting clear before you sit down with the numbers.
- Do you know your actual cost base for this year? Not the revenue. The cost of delivering what you sold. Payroll, tools, freelancers, software, marketing spend and the like. Before you can plan 2027 with any confidence, it helps to understand what 2026 actually costs to operate.
- What is AI costing you right now and what has it changed. Even a rough, qualitative answer is more useful than a blank. Ask the people in your team who are using these tools what is different about their work because of it. That conversation will tell you more than any subscription invoice.
- Does your paid ads conversation with clients reflect what the market actually looks like today? If you have not had it, the start of Q4 is a reasonable moment to do that, before next year’s budgets get set on assumptions that are two years out of date.
- Do you have an actual person who is in charge of AI spend-outcome in your business? If the answer is nobody in particular, that is fine to acknowledge. It is where most agencies are. But naming it means you can change it, and giving someone clear ownership of tracking what AI costs versus what it returns is one of the more useful things you can do before this year closes out.
The budget conversation does not need to be complicated. But it does need to actually happen with clarity.
Note: This article is meant for information only and not meant as advice. Please consult a professional for your specific case for clarification of doubts.
