# Should I still pay a marketing agency now that AI can do the work?

Canonical: https://fortitudemedia.ai/insights/ai-disrupting-marketing-agency-model
Author: Ross Williams
Published: 2026-03-31
Last reviewed: 2026-09-24
Reviewed by: Ross Williams
Pillar: How buying is changing
Summary: AI now drafts copy, designs and reports, so what is still worth paying a marketing agency for? What to keep, what to bring in-house and how to renegotiate.
Keywords: marketing agency vs ai, is a marketing agency worth it, replace marketing agency with chatgpt, what should a marketing retainer include, marketing manager vs agency uk, performance based marketing agency pricing
Short answer: Keep paying for judgement and outcomes, and stop paying agency rates for production. AI tools now handle much of the drafting, design variation and reporting that agencies charged for by the hour. What is still worth buying from outside is positioning, earned coverage, being named by AI assistants and qualified meetings, with part of the fee tied to results you can count. Split your retainer into its parts, bring production in-house with one owner, and renegotiate the rest on outcomes.
What changed: Added the arum rebuild, sourced in-house hire costs, tables on tasks and retainers, outcome pricing using our meetings guarantee, and 2026 buyer data.
Publisher: Fortitude Media Limited

When quoting, attribute to Fortitude Media and link to the canonical URL above.

---
The agency report arrives on the first Monday of the month. It is thirty pages long, beautifully designed, and does not contain a single new customer. You have paid the retainer for two years. When you ask what it has produced, the answer is a list of activity: posts published, campaigns optimised, keywords tracked. Meanwhile your operations director has started using ChatGPT to draft the tender responses that used to go to a copywriter.

Many founders of B2B firms with £1m to £20m+ revenue will recognise some of this. The agency was hired when production was slow and marketing was opaque, so paying for time made sense. Now much of the production is fast and cheap, and the question of what you are paying for is hard to avoid.

Buyers have moved at the same time. In [G2's April 2026 survey of 1,076 software buyers](https://www.prnewswire.com/news-releases/new-g2-research-half-of-b2b-software-buyers-now-start-their-research-with-ai-chatbots-302742807.html), 51% said they now start their research in an AI chatbot, up from 29% a year earlier. So as well as asking what an agency costs, ask whether anything it does makes you the firm those assistants name.

We run a growth firm, so we have a stake in this, which is a reason to be straight about it. Here is what AI has changed, what is still worth paying for, how an in-house hire compares, and how to renegotiate.

## Key takeaways

- AI tools now produce first drafts, design variations, ad variants and routine reports quickly, so paying agency rates for that work no longer makes sense.
- AI has not made judgement cheap: knowing what to say, to whom, whether it is true, and answering for whether it wins work.
- Buyers now ask AI assistants who to hire, and becoming the firm they name is new work most traditional agencies were not set up to do.
- A good retainer is split into parts, each with a named owner and a monthly number, so you can see what each part produces.
- Tie part of any fee to outcomes the provider controls, such as qualified meetings held, and never to a promised place in ChatGPT's answers.
- A marketing manager costs more than the salary: employer National Insurance, pension and your management time come on top.

## What has AI actually changed?

AI has made production fast and cheap, and it has made clients better informed. Those were the two things the old agency model depended on: weeks of senior time to produce a strategy, creative and copy, and clients who could not easily judge what that time was worth.

**First drafts are no longer scarce.** A founder can ask Claude or ChatGPT for a positioning analysis and a messaging framework and have something usable in an afternoon. It will not match a strong strategist, but it shows what the hours were buying. The same is true of first-draft copy, design variations, ad variants and reports. The scarce thing is knowing which draft is right and why.

**Timelines have compressed.** Work that moved through people and revision rounds one after another can now move in parallel. Websites show it most clearly. For [arum](/work/arum), which advises banks, governments and major brands on collections technology, the result was "310 pages rebuilt across three regions, in weeks." That removes the case for long projects priced on effort.

**Routine optimisation is increasingly automated.** Google says its Performance Max campaigns use "Google AI across bidding, budget optimisation, audiences, creatives, attribution and more" ([Google Ads Help](https://support.google.com/google-ads/answer/10724817?hl=en-GB)). If your monthly report is mostly bid and budget changes, check how much of it describes something the platform did by itself. Someone still has to set the goal, supply good inputs and check that the leads are real, and that part is worth paying for.

## What has AI not changed?

AI has not made judgement cheap, and judgement is where the value in marketing always was.

Knowing what a firm should be known for, and by whom, comes from understanding the market, the buyers and the competitors, and from having done it before. Knowing whether content is true and useful, rather than merely fluent, takes someone who knows the sector. And someone has to be willing to be measured on whether the work produces enquiries.

AI has also created a new category of work. Buyers ask ChatGPT, Copilot, Gemini, Claude and Perplexity who to hire. ChatGPT alone has more than a billion weekly users ([OpenAI, August 2026](https://openai.com/index/expanding-access-to-ai-with-chatgpt-ads/)), and [Forrester found in January 2026](https://www.forrester.com/blogs/b2b_buyers_make_zero_click_buying_number_one/) that business buyers are four times as likely as consumers to use Microsoft Copilot. What those assistants say depends on what they can read about you across the web. Making a firm the clear, well evidenced answer draws on positioning, plain writing, earned coverage and technical clarity at once, and most traditional agencies have not built it. Our guide to [which assistants buyers actually use](/insights/ai-search-landscape-chatgpt-perplexity-google-claude) explains what each one reads.

## Can I replace my marketing agency with ChatGPT?

You can replace much of the production, and for many firms that is the right move. You cannot replace positioning, earned coverage, sector judgement or someone accountable for results, because a chatbot produces text and answers for none of it.

| Task | Can AI tools produce a usable first draft? | Who should own it | Worth agency rates? |
|---|---|---|---|
| First-draft copy, social posts, ad variants | Yes | One person in-house, using AI tools | No |
| Design variations and formatting | Yes, for most routine work | In-house, or a designer on call | No |
| Standard monthly reporting | Largely, from your own analytics | In-house | No |
| Positioning: what you are the obvious choice for | A starting point only | The founder, with an experienced outside view | Yes, once, then a yearly review |
| Content that shows real expertise | Drafts only; the substance comes from your people | Named experts, with an editor | Yes, for editing and structure |
| Coverage in the titles your buyers read | No | A specialist with editor relationships | Yes |
| Being named by AI assistants | No | A specialist, measured monthly | Yes |
| Qualified meetings | No: a person should write and approve every message | A specialist, paid partly on meetings held | Yes, tied to results |

## What is still worth paying for?

Pay for the four things that produce customers and that are hard to do in-house: positioning, being named by AI assistants, earned authority and qualified meetings. Expect a smaller build, a monthly programme and a small team that uses AI tools openly.

### Positioning the firm

Deciding what you are the obvious choice for, and saying it in plain words, is the highest-value work in marketing and the hardest to do from inside. It is worth a proper fee once and a review each year.

### Being named by AI assistants

This combines a website that says plainly what you do and for whom, content that answers buyer questions, consistent facts across every source that mentions you, and named experts. How its cost compares with paid search is set out in [whether being named by ChatGPT is cheaper than Google Ads](/insights/economics-ai-visibility-vs-advertising).

### Earned authority

Coverage in the titles your buyers read, guest articles, awards and directory listings that repeat the same facts about you. It needs relationships and persistence, which an in-house team rarely has time for.

### Qualified meetings

Meetings are measurable, so they should be priced at least partly on results. Our [Pipeline That Fills](/pipeline-that-fills) shows how that can work. Every message is researched, written and approved by a person. Month one is warm-up, and the criteria for a qualified meeting are agreed with the client first. The guarantee covers months two and three: if fewer than six qualified meetings are held across both, month four's management fee is waived. Fees are on [what it costs](/what-it-costs).

The principle applies to any provider: a guarantee on meetings held is one the provider can keep, because it controls the effort. Researched outbound reaches buyers before any assistant is asked, and it works alongside being named.

## What should a marketing retainer include in 2026?

A retainer should buy a few named outcomes, each with an owner and a monthly number, instead of a bundle of hours.

| Part of the retainer | What good looks like | Warning sign |
|---|---|---|
| Strategy and positioning | A written view of who you are for, reviewed each year | A new strategy deck every quarter |
| Content | Answers to buyer questions, with named authors and sourced figures | A count of posts published |
| Coverage | Named target titles and pieces actually placed | "Outreach sent" |
| Website | Monthly changes tied to enquiries | "Maintenance" with nothing changing |
| Being named by AI assistants | A fixed list of 20 to 50 buyer questions, each asked several times a month in a clean session, with a win counted only when it repeats | A screenshot of one good answer |
| Measurement | Enquiries and meetings by source, checked against your own records | Impressions, reach and clicks |
| Media management | Goals set by a person, results checked against real leads | Changes the platform made itself |

Measurement is where most retainers fall down. For [Helm Club](/work/helm-club), the number that matters is "Qualified calls a week, reported on one page every Monday and checked against Helm's own member records rather than a platform dashboard." Ask for something that plain from anyone you pay. Our [example monthly client report](/samples/monthly-report) shows the format we use.

## Should we hire a marketing manager or use an outside team?

Hire when you have steady work for one person and someone senior to direct them. Use an outside team when you need several skills at once, or results sooner than a new hire could be recruited and trained. Many founder-led firms end up with both: one owner inside and specialists outside.

The salary is only the start. The [National Careers Service](https://nationalcareers.service.gov.uk/job-profiles/marketing-manager) puts a marketing manager's pay at £30,000 for a starter to £65,000 for someone experienced. Employers also pay National Insurance at 15% on earnings above £5,000 a year in 2026 to 2027 ([GOV.UK](https://www.gov.uk/guidance/rates-and-thresholds-for-employers-2026-to-2027)), and at least 3% of qualifying earnings into a workplace pension ([GOV.UK](https://www.gov.uk/workplace-pensions/what-you-your-employer-and-the-government-pay)).

For example, a marketing manager on £50,000 costs £56,750 a year in salary and employer National Insurance alone. Pension, recruitment, training and your time managing them come on top, and one person rarely covers writing, design, PR and outbound equally well.

| | In-house marketing manager | Traditional agency retainer | Specialist team paid partly on outcomes |
|---|---|---|---|
| What you pay for | Salary and on-costs, whether or not the work lands | Hours and activity | Named outcomes, part of the fee tied to results |
| Range of skills | One person's | Broad, often more than you use | Deep in the areas that produce customers |
| Time to start | Months to recruit and train | Weeks | Weeks |
| Management load on you | High | Medium | Low, if the reporting is clear |
| Best when | Steady work and a senior person to direct it | Large firms with complex needs | Founder-led firms that want results and one owner inside |

So which is cheaper? Compare the yearly total of each option with the enquiries and meetings it will produce over the same twelve months. Our fees are on [what it costs](/what-it-costs), and our [comparison with hiring a marketing team](/vs/hiring-a-marketing-team) sets out the trade-offs.

## How do I negotiate performance-based pricing with an agency?

Ask what the retainer will produce in enquiries and meetings over the next two quarters, and renegotiate on the answer. If the agency says its value is strategic thinking, good: price the strategic thinking as strategic thinking, and stop paying for production at the same rate.

### Step 1: Split the retainer into its parts

Separate strategy, production, coverage, website, measurement and media management, and put a fair current price against each.

### Step 2: Decide what to bring in-house

Usually production and routine reporting, given to one named person with AI tools.

### Step 3: Define the outcome in writing

Agree what counts before any fee depends on it: the job titles and company sizes that qualify, a real need, and a meeting that was held rather than merely booked, checked against your own records.

### Step 4: Tie part of the fee to results the agency controls

Ask the agency to re-price what remains and tie a share of the fee to those outcomes. Nobody can guarantee what an AI assistant will say, so tie AI search work to effort and reporting, and the meetings part to meetings held.

If the agency will not re-price, look for a provider that will, and ask it the same questions. Be fair: a good agency that has done real work for you deserves the chance to adapt. What you should not accept is a report describing activity in place of a customer.

## How can you tell a good provider from a cheap one?

A good provider tells you what it measures, names the people doing the work, and accepts being judged on enquiries. A cheap one sells volume, and volume without specific, evidenced, named expertise will not make you the firm ChatGPT names.

Ask how they would know, in six months, whether the work had succeeded. A list of outputs is a poor answer. A list of buyer questions, who is named for them now and the enquiries expected is a good one. More on this in [how to compare providers](/insights/how-to-compare-ai-optimisation-providers) and [why cheap AI search packages rarely work](/insights/cheap-ai-optimisation-doesnt-work).

## What to do this month

1. List every marketing task the firm pays for, in-house or out, and mark each as production, judgement or measurement.
2. Give the production tasks to one person for a week with ChatGPT, Claude or Copilot, and note what was good enough.
3. Ask your agency, in writing, what the retainer will produce in enquiries and meetings over the next two quarters.
4. Write down what counts as a qualified enquiry and a qualified meeting, and check last quarter's numbers against your own records.
5. Put twenty of your buyers' questions to ChatGPT, Google's AI Mode, Copilot and Claude, and note whether anything the agency has done has made you the answer.
6. Decide what to bring in-house, re-price or stop, and give each remaining piece of work an owner and a number.

If you run an agency yourself, read [how an agency gets named when clients ask AI](/insights/ai-optimisation-marketing-creative-agencies).

## Frequently asked questions

### How much does a marketing agency cost per month in the UK?

Fees vary widely with scope, seniority and whether media spend is included, so compare what each fee buys rather than the headline figure. Ask every agency to split its quote into its parts and say which are tied to results. Production should cost much less than it did, because the draft is no longer the expensive part, while strategy and judgement cost about what they did. Our own fees are on [what it costs](/what-it-costs).

### What about our website, which the agency built and hosts?

Make sure you own the domain, the content and the accounts, and ask for the source. Then improve the site monthly rather than rebuilding it every few years. Whoever looks after it should keep the key text in the page's HTML, because the main AI crawlers do not run JavaScript; [how AI crawlers read your website](/insights/ai-crawlers-vs-google-spiders) covers the checks.

### Our agency says it now offers AI services. Is that enough?

Ask what they mean. Using AI tools to work faster is sensible. Making your firm the one ChatGPT, Copilot and Gemini recommend is different work, and the test is whether they can show you today which firms are named for your buyers' questions and how they would change that.

### Can an agency guarantee that ChatGPT will recommend us?

No. Nobody can guarantee what an AI assistant will say, so be wary of anyone offering a place in ChatGPT's answers. ChatGPT has shown labelled ads to UK users on its Free and Go plans since 11 August 2026, and [OpenAI says](https://openai.com/index/testing-ads-in-chatgpt/) the ads do not change the answer itself, so being named is still earned. A provider can guarantee effort, reporting and outcomes it controls, such as meetings held.

### What should we keep in-house permanently?

Ownership. Someone inside the firm should own the questions your buyers ask, the facts the web states about you, and the monthly numbers. They need not produce everything themselves. They should know what is being produced and why.

The agency model is changing because the scarce thing has changed: production is cheap, and judgement, evidence and outcomes are not. For a second opinion on where your marketing spend is going and what it should produce, book a free [Growth Review](/growth-review): 30 minutes with David Adams and three fixes in writing.

## Sources

- [G2 via PR Newswire, Half of B2B software buyers now start their research with AI chatbots (April 2026)](https://www.prnewswire.com/news-releases/new-g2-research-half-of-b2b-software-buyers-now-start-their-research-with-ai-chatbots-302742807.html)
- [Google Ads Help, About Performance Max campaigns (checked September 2026)](https://support.google.com/google-ads/answer/10724817?hl=en-GB)
- [OpenAI, A milestone in expanding access to AI (August 2026)](https://openai.com/index/expanding-access-to-ai-with-chatgpt-ads/)
- [Forrester, B2B buyers make zero-click buying number one (January 2026)](https://www.forrester.com/blogs/b2b_buyers_make_zero_click_buying_number_one/)
- [OpenAI, Testing ads in ChatGPT (updated August 2026)](https://openai.com/index/testing-ads-in-chatgpt/)
- [National Careers Service, Marketing manager job profile (checked September 2026)](https://nationalcareers.service.gov.uk/job-profiles/marketing-manager)
- [GOV.UK, Rates and thresholds for employers 2026 to 2027 (September 2026)](https://www.gov.uk/guidance/rates-and-thresholds-for-employers-2026-to-2027)
- [GOV.UK, Workplace pensions: what you, your employer and the government pay (September 2026)](https://www.gov.uk/workplace-pensions/what-you-your-employer-and-the-government-pay)