Most marketing teams use AI now. In Salesforce's latest survey of about 4,450 marketers, 75% use it in some form, but only 13% use agents that run work on their own (Salesforce). Gartner expects the share of marketing work AI automates to more than double, from 16% today to 36% by 2028. Using AI was the first stage of this evolution. Automating with it is the next one, and most teams haven't started.
We hear the shift from our own clients. A year ago the questions were about which AI tool to buy. Now they ask about skills, plugins and automation, and they want whatever we build handed over with their team trained to run it. That puts a fair question on every marketing leader's desk: where is my team, really?
Nobody is as far along as they say
Ask a marketing leader how far along their team is with AI, and the answer usually sounds better than the reality. AI shows up in almost every task now, so it feels like the work has changed. The numbers tell a different story, especially for smaller teams.
That gap is an opening. The bar rises every quarter, and very few teams are clearing it yet.
Why most teams stall
Most teams bought tools and kept the old workflow. Drafts got faster. The approvals and handoffs stayed the same, so the results did too.
BCG found AI leaders spend about 70% of their effort on people and process, 20% on technology and data, and 10% on algorithms (BCG). Every major framework we reviewed, from McKinsey and Deloitte to Gartner and Microsoft, scores the same areas, with governance running through all of it. Most teams have invested in one and skipped the other two.
People
Roles, ownership and the AI fluency bar.
Process
Workflows, approvals and human checkpoints.
Tech
A shared workspace, agents and an agent-ready site.
The three legs depend on each other. A redesigned workflow stalls without people who own it, and both stall on a platform that can't keep up.
People: the team is being rebuilt now
Marketing teams are changing shape. Leaders are rewriting job descriptions and adding agents alongside people, and nobody has a finished model to copy. Everyone is working out what the team of the next few years looks like while still running this quarter's campaigns. What is already clear is that the change has started, and it is moving faster than most teams' plans for it.
18%
of marketing leaders have cut roles because of AI. 16% have created new ones.
Gartner via Marketing Dive2030
Most high-performing teams will have cut traditional entry-level roles by then.
Gartner via Marketing DiveSpecialists are giving way to generalists who direct AI across several areas, at companies of every size. AI fluency is a job requirement now. Shopify added AI use to performance reviews, and Zapier holds every employee to a baseline fluency level, which we saw firsthand at Webflow Conf. Waiting for the end state to become clear is the riskiest plan on the table. A year from now, the teams that waited will be building the team their competitors already have.
| Today | Next stage | |
|---|---|---|
| AI baseline | Some people, on personal accounts | Everyone weekly, on one team account |
| Ownership | Nobody owns the automations | Each skill and agent has a named owner |
| Job descriptions | Written around tasks | Written around outcomes and directing AI |
| Training | A one-time tool demo | Ongoing coaching and learning time |
| Leadership | AI handed off to the team | Leaders use AI in the open |
Process: redesign the workflow, not the task
McKinsey and Deloitte land in the same place: the unit of change is the end-to-end workflow. Speeding up one step, like copywriting, barely moves the campaign. Publicis Sapient learned this the hard way. Its first six months of “just use AI” changed very little, because the workflows stayed the same.
The teams that get there map every recurring task as AI-led, AI-assisted or human-only, and decide which steps stay human before automating anything. They start with one workflow and a small team, prove it, then expand in waves. They put checkpoints where judgment matters and cut approval layers that exist out of habit. Three of the workflows below run through the website, which is where the tech question starts.
| Today | Redesigned | |
|---|---|---|
| Weekly site report | Pull analytics, build slides | An agent flags drops, posts to Slack, opens tasks |
| Landing pages | Brief, design, dev ticket, days of waiting | Built from the design system in hours |
| On-brand copy | Rewrite until it sounds like us | A brand-voice skill drafts it right the first time |
| Content refresh | An annual audit | An agent flags decaying pages and drafts updates |
| Campaign recaps | Hours of compiling | A skill synthesizes results in minutes |
Tech: one system for people and agents
Gartner names the problem directly: CMOs “invest in AI tools faster than they build the data foundations, processes, governance and talent required to scale them.” The stack behind agentic marketing has four layers:
- A shared AI workspace, such as a Claude Team account, where everyone works from the same skills, context and brand rules.
- One place to run marketing work with agents. Webflow's new Source, now in limited research preview, gives marketers, designers and developers their own views, and agents follow the same permissions and approvals as people.
- An agent-ready website. On Webflow, MCP connects Claude, ChatGPT and Cursor to the site, Agent Instructions hold the brand and the rules, and every agent action is logged.
- Connected data, because agents only know what they can see.
93%
of marketing leaders still need a developer or agency to change their site.
Webflow 2026 State of the Website62%
of technical leaders blame the platform for their AI barriers.
Webflow 2026 State of the WebsiteThe website is where most stacks break. A team can have the right people and a redesigned process and still stall if a page change waits on a ticket. The platform is the ceiling.
| Today | Next stage | |
|---|---|---|
| AI workspace | A dozen personal accounts | One team account with shared skills |
| Where AI works | A separate chat window | Inside Slack and the tools you already use |
| Brand guidelines | A PDF on a shared drive | A skill every draft starts from |
| Website | Changes wait on a developer | Marketers and agents ship, with review |
| Data | Stuck inside each tool | Connected, so agents can see results |
5 easy ways to get closer to agentic marketing
The next stage puts agents in every conversation where your people work, all managed in one place. Getting there starts with small moves. Each of these takes days to start, and each one moves at least one leg forward.
Map your recurring work, then mark what AI can take.
List every weekly task and tag it AI-led, AI-assisted or human-only. That list becomes your automation plan.
Read: Seer Interactive's five stages of AI marketing maturity
Move your team to one AI account, then share everything.
A team account, such as Claude Team, gives everyone the same skills, context and brand rules.
Add AI where your team already works, starting with Slack.
No new app to open. AI joins the conversation.
Turn your brand guidelines into a skill, then use it everywhere.
Every draft, from anyone on the team, starts on-brand.
Automate one weekly task first, then move on to the next.
The weekly site report is a good first win. Each quick win makes the case for the next one.
Where to start: the first 90 days
- Find your real starting point. For each leg, ask where your team sits today: playing with AI, using it daily, automating with it, or running agents. Most teams land a stage lower than they expect. The lowest of the three is your real stage, because that leg will hold the other two back.
- Set the baseline. Move everyone onto a shared team account and expect weekly AI use from the whole team. Leaders go first and show their own work.
- Pick one workflow and give it an owner. Choose something recurring, high-volume and low-risk, like the weekly site report or campaign recaps. One named person owns the result, including whatever the AI produces.
- Map it and decide what stays human. Write down every step, then mark each one as AI-led, AI-assisted or human-only. Strategy, the original idea and final brand sign-off usually stay with people.
- Redesign it end to end. Build the skill, automate the steps that repeat, add checkpoints where judgment matters, and cut at least one approval that exists out of habit.
- Check the platform. If the workflow touches the website and every change still needs a developer, that is your ceiling. Fix it before you scale.
- Measure and share what changed. Track time saved, speed to launch and pipeline against how the workflow ran before. Show leadership the numbers.
- Repeat in waves. Take what you learned into the next two or three workflows. Each round goes faster than the last.
Deloitte found that AI pilots built with a strategic partner are twice as likely to reach full deployment as those built alone (Deloitte). That is the work we do. We build the skills your team runs on, hand them over, train your people to run them, and build your site on Webflow so agents can do their part.
