AI marketing agents built around your tools and your approval rules
An AI marketing agent plans and carries out marketing tasks through your tools, within limits you set. We build custom agents for lead qualification, content, paid media, email and reporting, or adapt a pre-trained one from our Bot Store.
Updated
Works inside your stack
Connects to your CRM, email platform and ad accounts through their APIs. There is nothing to replace.
A person approves what goes out
Customer-facing messages and budget changes can wait for approval before anything happens.
Engineers from $45 an hour
Published rates, and a written quote after discovery for a defined project.
What is an AI marketing agent?
An AI marketing agent is software that uses a language model to work towards a marketing goal, deciding the next step and carrying it out through your tools. It might research a new lead, reply, book a call and update the CRM, while a person approves anything that carries risk.
The difference from an ordinary marketing tool is who decides the next step. An automation runs the sequence you drew. An agent reads the situation, picks from the tools it has been given, looks at the result and decides what to do next, inside limits you set.
Take a new enquiry from your website. The agent replies within seconds, looks up the company, asks the questions your sales team would ask, scores the lead against your ideal customer profile, books a meeting in the right calendar and logs every step in your CRM. If the lead is outside your profile, or asks for a person, it passes the conversation on with what it has learned so far. Our pre-trained AI SDR agent works this way.
What can AI agents do for a marketing team?
The repetitive work between a decision and its result: replying to and qualifying leads, drafting content from approved sources, checking ad accounts, running follow-up journeys and pulling reports together. Your team keeps the strategy, the brand and the final say on what reaches customers.
Swipe the table sideways to see every column.
| Agent | What it does | Connects to | What a person checks |
|---|---|---|---|
| Lead qualification and SDR | Replies to enquiries in seconds, researches the company, asks qualifying questions and books meetings | CRM, calendar, web forms, WhatsApp | The handover, and any outbound message you choose to review |
| Content and SEO | Drafts briefs, pages and product copy from your approved sources, and checks them against your brand rules | CMS, keyword data, your documents | An editor approves everything before it is published |
| Paid media operations | Pulls performance across accounts, flags wasted spend and drafts ad variants and budget changes | Google Ads, Meta, analytics | A person approves every budget or bid change |
| Lifecycle email and CRM | Builds segments, drafts journeys and follows up abandoned quotes and bookings | Email platform, CRM, customer data platform | Templates and offers, with consent and frequency rules applied to every send |
| Reporting | Collects data across channels, explains what changed and drafts the weekly report | Analytics, ad accounts, CRM | An analyst checks the figures before they are shared |
These are typical set-ups, agreed for your business before a build. Results depend on your data, your channels and your offer, so we set success criteria with you rather than quoting a percentage.
Lead work is a sensible place to start. In the European Marketing Agenda 2026, 49% of respondents expected lead generation and qualification to be among the disciplines that benefit most from agentic AI (CIM). It also has a clear result to measure: how many qualified meetings reach your sales team, and how fast. See AI lead qualification.
Our rates
Published prices in US dollars. Monthly plans are 160 hours, about 15% below the hourly rate.
Mid-level developer
3 to 5 years
$25 / hour
$3,400 a month (160 h)
Senior developer
6 or more years
$35 / hour
$4,750 a month (160 h)
AI / LLM engineer
Agents, RAG, IBM watsonx
$45 / hour
$6,100 a month (160 h)
Senior AI engineer
Leads AI architecture and delivery
$60 / hour
$8,150 a month (160 h)
Two-week trial with the developer you choose. No charge if you stop within the first five working days. US Eastern clients get 09:00 to 12:00 US Eastern time through an evening shift, at 15% extra. Full pricing and comparisons.
Which marketing job should an AI agent take on first?
One that repeats often, has a clear result and costs little if a draft is wrong. Replying to new leads and following up abandoned quotes usually fit. Writing your brand campaign, or spending your budget without review, do not, at least not at first.
| Job | Why it fits, or doesn't | How to measure it |
|---|---|---|
| First reply and lead qualification | Happens many times a day, and speed matters | Time to first reply; qualified meetings booked |
| Abandoned quote or booking follow-up | A clear trigger and a clear outcome | Quotes or bookings recovered |
| Weekly performance reporting | Repetitive, and a person still checks the figures | Hours saved; errors found in review |
| Product and landing page drafts | Works from your approved facts, and an editor reviews it | Editing time per page |
| Ad budget changes | Start with recommendations only, because mistakes cost money | Recommendations accepted by your team |
| Brand campaigns and positioning | Keep these with your team: judgement matters more than speed | Not a good first job |
How is an AI marketing agent different from automation, a chatbot or a copilot?
Automation follows a fixed sequence. A chatbot answers questions. A copilot helps one person work faster and waits to be asked. An agent takes a goal, chooses the steps itself and acts through your tools within set limits, so the work continues without someone starting each step.
Swipe the table sideways to see every column.
| Tool | What starts it | What it can do | Best for |
|---|---|---|---|
| Marketing automation | A trigger you defined | Runs the exact sequence you drew | Journeys that rarely change |
| Chatbot | A visitor's question | Answers from a script or your documents | FAQs and first-line answers |
| Copilot | A person asking for help | Drafts or summarises for that person | Speeding up one person's work |
| AI marketing agent | A goal, a new event or a schedule | Chooses the steps and acts through your tools, within limits | Work that varies case by case and has a clear result |
Many teams need less than an agent. If your follow-up sequence works and rarely changes, leave it as automation. Our guide to AI agents vs chatbots vs RPA explains how to tell the difference.
Should you use your platform's built-in agents or build a custom one?
Start with what your marketing platform already includes, because it is quick and often enough for work inside that one platform. Build a custom agent when the job crosses several systems, depends on your own data and rules, or needs approval steps and ownership the platform does not give you.
| What to compare | Built-in platform agent | Custom AI marketing agent |
|---|---|---|
| Works across | Mostly its own platform | Every system you connect, including your own |
| Your rules | The settings the platform offers | Your approval steps, limits and brand rules, written into the build |
| Speed to start | Often days | Typically six to twelve weeks for a first agent |
| Cost | Licence or usage fees | A one-off build plus model usage and maintenance |
| Who owns it | The platform | You: the code, prompts and test sets live in your repositories |
| Best for | Work inside one platform | Work across systems that the platform's agents do not reach |
Many teams are still at the start of this. A large 2025 survey of 4,450 marketers found that 75% of marketing organisations use AI, but only 13% use agentic AI (survey). Gartner reported in October 2025 that 45% of martech leaders with agents in pilot or production said vendor-offered agents do not meet their expectations (Gartner). So test a built-in agent against your own success criteria before you pay for a custom one, and build only where it falls short.
Our guide to custom agents vs off-the-shelf chatbots goes through the costs of each.
How do you build a custom AI marketing agent?
One job at a time. Pick a task with a clear result, connect the tools it needs, write down its limits and brand rules, test it against your past campaigns and enquiries, then launch it with a person approving its output. Approval steps are relaxed only where the results earn it.
- 1
Pick one job
Choose a task with a measurable result, such as qualified meetings booked or quotes recovered, and write down what success looks like.
- 2
Connect the tools
Give the agent access to the CRM, email platform or ad account it needs through their APIs, with the least access that works.
- 3
Set the limits
Write down what it may do alone, what needs approval, the brand voice, the claims it must not make, and the consent and frequency rules.
- 4
Test on real history
Score it against past enquiries and campaigns, and against hostile inputs such as prompt injection, before any customer sees it.
- 5
Launch with approvals
A person approves its output at first. Each action is logged, so you can see what it did and why.
- 6
Measure and widen
Compare results with your success criteria. Relax approvals where it has earned that, then add the next job.
How we build sets out our delivery method, and how long an AI agent takes to build covers the timeline.
What data does an AI marketing agent need?
Access to the systems it works in, your approved facts and examples of good work. That usually means your CRM and email or ad accounts, a brand guide, product and price information, and a sample of past enquiries or campaigns to test against. It does not need your whole database.
- System access: accounts in your company's name, with the least access the job needs.
- Approved facts: products, prices, offers and the claims you allow, kept current by someone on your team.
- Brand guide: your tone of voice, with examples of copy you were happy with.
- Test examples: a sample of past enquiries or campaigns. Two to three weeks of clean examples is usually enough to start.
- Consent records: who has agreed to hear from you, on which channel, so the agent can respect them.
Whether a hosted model provider may train on your data depends on its terms and your account settings, which we check with you first. If customer data must not leave your network, the agent can run on an open-weight model on your own servers: see private AI.
How do you keep a marketing agent on brand and within the rules?
By limiting what it can say and do, then checking the result. The agent works from your approved brand guide, claims and templates. Consent and frequency rules apply to every message, risky actions wait for a person, and every action is logged so your team can review it.
- Brand voice: it writes from your guide and past approved copy, and is tested on them before launch.
- Claims: it uses only the facts, prices and offers you have approved, so it cannot invent a discount.
- Regulated sectors: pre-approved templates and mandated disclosures, which is how our banking and insurance marketing agents work.
- Consent: it respects opt-outs and frequency caps, and sends nothing to anyone who has not agreed.
- Spend: it cannot change a budget or a bid without the approval you set.
- Disclosure: customers are told they are talking to an AI and can ask for a person at any point.
Your own legal and compliance teams decide what your rules should be. We build them into the agent and test them. Our security page covers data handling and contracts.
Do you have pre-trained marketing agents for specific industries?
Yes. The Bot Store has marketing agents for banking, insurance, airlines, hotels, travel and vehicle rental, plus an AI SDR agent for any industry. Adapting one to your systems is usually faster than starting from nothing, and we will say whether one is close enough.
AI SDR agent
Replies to inbound leads, qualifies them and books meetings.
→ Read more
Retail banking
Segments, compliant offer copy and churn prediction.
→ Read more
Credit cards
Acquisition, activation and attrition campaigns.
→ Read more
Corporate banking
Account-based campaigns and lead scoring for relationship managers.
→ Read more
Health insurance
Lead capture, quote recovery and renewal campaigns.
→ Read more
Airlines
Personalised offers and abandoned booking recovery.
→ Read more
Hotels
Direct bookings, soft-date offers and guest win-back.
→ Read more
Travel
Abandoned search recovery and seasonal packages.
→ Read more
Car and truck rental
Idle-fleet offers, quote recovery and add-ons.
→ Read more
What does an AI marketing agent cost, and how long does it take?
A first agent scoped to one task with one or two integrations typically takes six to twelve weeks. By our planning estimates, an agent that takes actions needs 800 to 1,600 hours, or $42,000 to $84,000 at our rates. Engineering is $45 to $60 an hour, and model usage is billed by your provider.
A simpler assistant that only reads your systems and drafts for a person costs less: 480 to 960 hours, or $22,800 to $45,600 by the same estimates. These are planning estimates, not quotes. A defined project is quoted in writing after discovery, before any build starts. Our AI agent cost guide shows what moves the figure up or down.
Running costs come on top: model usage, hosting in your own account, and maintenance time to review its work and update its rules. Every rate we charge is on our pricing page.
Can a marketing agency offer these to its clients?
Yes. Under our white-label program we build and deliver the agents behind the scenes while your agency keeps the client relationship and sets the price. We sign your NDA before the first call, and we never contact your clients unless you ask us to.
See the white-label program for how it works, or developers for agencies if you would rather add engineers to your own team.
Frequently asked questions
Will an AI marketing agent replace my marketing team?
No. It takes on repetitive work such as first replies, data pulls and follow-ups. Your team keeps the strategy, the brand and the decisions, and approves what you decide needs approving.
Do we need a new marketing platform?
No. The agent works with the CRM, email platform and ad accounts you already use, and writes its notes and actions back into them.
Can it work with HubSpot or Salesforce?
Yes. Both can be connected through their APIs, as can other CRMs such as Zoho and Microsoft Dynamics. Each action is a separate tool with its own permissions, tested on test data first.
Can an AI agent run our ads on its own?
It can, but we do not recommend starting there. Begin with reporting and draft recommendations, and keep a person approving budget and bid changes until the agent has earned more trust on your own accounts.
Can it write in our brand voice?
Yes. It works from your brand guide and past approved copy, and it is tested against them before launch. An editor approves content until you are happy to relax that.
Will AI-written content hurt our SEO?
Google's concern is whether content is useful, not whether AI helped write it, but mass-producing thin pages to game rankings breaks its spam policies. Use the agent for research and first drafts from your own facts, and keep a person editing what is published.
How do you stop it messaging people too often?
Frequency caps and consent rules apply to every message, and the agent can see the customer's history before it sends anything. People who opt out stay out.
How do we measure whether it is working?
Against the success criteria agreed before the build, such as qualified meetings booked, quotes recovered or hours saved on reporting. The pilot ends with a report that includes the failures, and you decide whether to go further.
What is the difference between an AI marketing agent and an AI marketing agency?
An AI marketing agent is software. An AI marketing agency is a company that runs marketing for you using AI tools. We build the software, and we also offer digital marketing services if you want people to run the campaigns.
Talk to an engineer about your marketing agent
A conversation about your goals and systems, and a straight answer on whether an agent is the right tool. Paid work starts only after you have agreed a written scope.
- We sign your NDA before the first call if you need one
- You leave with a next step, even if the answer is "you don't need an agent"
- The code and IP are yours if you go ahead
Book a free first call
Pick a time, or send us the details and we will reply.