Getting started

Where to Start With AI in Your Marketing

Four places AI earns its keep in a small business's marketing: follow-ups, content, reporting and lead handling. What each takes, and where to start.

By Ejaz Ahmed, Founder, AI Marketing Foundry · Published

· Last reviewed

·

8

min read

var(--variable-VlAYQqBdT)

Most writing about AI in marketing fails in one of two directions. It is either too high up to act on, or it is a Python script. Neither helps if you run a business, or a marketing team, and want to know where to spend the next fortnight.

So this is the middle version: four marketing systems a UK business can realistically set up, all of which I have either built for clients or run myself. It is also the starting point for everything else on this blog, so if you are working out where to begin with AI in your marketing, begin here.

One note on the figures before you read them. The costs below are my own estimates from building these systems, not quoted prices. Vendor pricing changes constantly and usually scales with volume, seats or list size, so treat these as planning ranges and take the real number from each vendor's own page on the day you buy.

1. A lead qualification and routing system

This is the highest return system most SMEs can build, because it works on the part of the funnel you have already paid for.

When someone submits an enquiry form, a connected system adds what it can find about them (company size, sector, LinkedIn profile), scores them against criteria you have defined, and routes the qualified ones straight to your calendar. Unqualified enquiries go into a nurture sequence, and the ones that need more context get a specific follow-up instead of sitting in an inbox.

You stop spending your best hours on enquiries that were never going to convert, and the sorting happens whether or not you are at your desk.

What it takes: a form tool such as Typeform or Tally, a CRM or Airtable, an enrichment layer such as Clay, and automation glue such as Make or Zapier. My estimate, from builds of this shape: roughly £100 to £200 a month to run, and one to two days to set up properly including testing. Enrichment is usually the line that moves, because it is priced per record.

2. Automated follow-up sequences

Small businesses lose a lot of deals in the follow-up rather than the pitch. Often the client never went cold. A busy fortnight happened and nobody chased.

A sequence runs regardless. Someone enquires on a Wednesday evening and gets a considered response within the hour. No reply after three days brings a second email. Still nothing a week later brings a different angle. The copy has to be good, and writing it is the actual work, but you write it once.

The best result I have on this is a fashion client's email programme that returned 5.5x over twelve months at a 47% average open rate. When I want to know why a programme like that is performing, I query the account directly from Claude rather than exporting spreadsheets, which I wrote up in Klaviyo MCP and Claude.

What it takes: Klaviyo, ActiveCampaign or HubSpot, an honest view of your typical sales timeline, and good copy. My estimate: around £30 to £100 a month at small list sizes, and about a day to build the first sequence. Email platforms price by contacts, so this is the figure most likely to rise as you grow.

3. An AI-assisted content pipeline

Producing content consistently is hard when you are also running the business. Most SMEs either do it in bursts or hand it to an agency at four figures a month.

There is a middle path. You supply the ideas and the angles, AI produces first drafts, and you edit and publish. You still edit everything, but it is fast enough to produce eight to twelve pieces a month on a few hours of real work.

I run this for my own content and for clients. On one personal brand, I took it further and built a custom AI web app that handles the whole weekly pipeline: what used to take about twelve hours a week now takes about ten minutes of my time. That number depends entirely on the setup work, which was writing down how I write, what I will not say, and what finished looks like. Without that document, the same tools produce filler.

Build in one caveat from the start: search is sending fewer clicks to sites than it used to, so a pipeline that only produces search-shaped blog posts is aiming at a shrinking target. Plan for the content to work in email, on social and in front of people who already know you, as well as in a results page.

What it takes: Claude or ChatGPT, a content calendar, and a written brand voice document. My estimate: the assistant subscription plus two to three hours a month of oversight. Claude Pro, the tier I use, is listed at $20 a month on Anthropic's own pricing page (checked August 2026), shown in local currency where supported.

4. A paid media reporting system

If you run Google Ads or Meta Ads, reporting quietly eats a day a month, and explaining performance to someone else eats more.

A connected setup pulls data from the ad platforms, applies your own performance thresholds, and flags what has moved outside them. Add AI-generated commentary and most of the report writes itself.

The bigger gain is in the analysis. The Google Ads account I am proudest of was built from scratch and runs at 8.63x ROAS, with AI doing the planning support, the audits and the optimisation legwork while I made the decisions. Reading a search terms report line by line used to be a job I postponed to next month. Doing it the same afternoon is where the return came from, because in paid media a postponed audit is a month of wasted budget.

What it takes: a data aggregator such as Supermetrics or Funnel.io, and a visualisation layer such as Looker Studio or a custom HTML dashboard. My estimate: roughly £50 to £150 a month, driven mostly by how many data sources you connect.

Which one first

You do not build all four. You build the one that matches your current bottleneck, and you leave the rest alone until it is working.

  • Enquiries arrive but go cold before you reply: build the follow-up sequences first. Fastest to build, and it works on demand you have already generated.

  • You get plenty of enquiries and most are wrong for you: build lead qualification and routing.

  • You go quiet for months at a time because nobody has time to write: build the content pipeline.

  • You are spending real money on ads and cannot say what is working: build the reporting system, then use it to run the audit you have been avoiding.

Each of these is a project rather than a product you buy. They need setup, testing and a few rounds of correction, and the ones that hold up are built around how your business works rather than copied from a template. What makes them last is knowing how to brief, check and fix them, and a team can learn that.

Turn this into a system

This method is part of the Getting started module in our training. Your team builds it on your own marketing and keeps it.

See the training

Frequently asked questions

What can AI actually do for a small business in the UK?

Four things worth building: qualify and route inbound leads, run follow-up sequences without manual chasing, produce content drafts in minutes, and turn ad platform data into a report and an audit. All four are buildable without an enterprise budget.

How much does it cost to implement AI in a small business?

The figures in this article are my own planning estimates from building these systems, not published prices. Broadly: lead qualification around £100 to £200 a month, email automation £30 to £100 depending on list size, an AI content setup at the cost of one assistant subscription, and paid media reporting £50 to £150. Check each vendor's current pricing before you commit, because these move.

How long does it take to build an AI system for my business?

In my experience, one to two days for the first working version, including testing. Deciding the rules the system runs on, such as what counts as a qualified lead or an acceptable cost per acquisition, takes longer than the build.

Do I need a developer to build these AI systems?

Not usually. Make, Zapier, Klaviyo and Airtable are no-code, and Clay needs configuration rather than coding. You need somebody comfortable setting up software carefully, not a development team.

Where should a UK small business start with AI?

With whichever bottleneck is costing you most right now, using the four cases above. Build one, get it working properly, then move on. Running two half-finished systems is worse than running one that works.

About the author

Ejaz Ahmed founded AI Marketing Foundry after six years in performance marketing at WPP and Publicis agencies. He holds an MSc in Data Science and AI and teaches teams to use AI across their marketing, with systems they keep running.

You’ve got the tools. Let’s build the method.