6 Lessons for Building an AI-First Marketing Strategy (from Vendavo's CMO)

An AI-first marketing strategy starts with the buyer and the data, not the tool list. Start small, fix your data sources, keep human judgment on the decisions that matter, and put an agent where buyers already look for answers.

That is the TLDR version of what Vendavo's CMO, Lindsay Duran, shared with Docket's Head of Marketing, Lauren McHugh in a recent webinar. Vendavo is an AI-first company, and its marketing team is building around AI in the open. They have wins to share, and they are honest about what has not gone to plan.

Here are six lessons from the conversation, plus what Vendavo learned from putting an AI agent at the front of its website.

Prefer to watch? Watch the full webinar on demand.

In this post

1. Start now, and fix your data first
2. Give AI specific jobs, not vague ones
3. Keep judgment human, and expect review work
4. Measure pipeline, not labels
5. Know where humans stay in the loop
6. Put the agent where buyers ask questions

1. Start now, and fix your data first

Lindsay's first piece of advice for CMOs is simple: don't be afraid to get started. When Vendavo began rolling out its intelligence platform, built on Claude, she told her team that everyone needed to complete training on how to create skills by a set date. There was a lot of testing and building ahead, so waiting was not an option.

Her second piece of advice is where most teams stall. Think hard about your data sources, AI produces output at scale, so if your sources are out of date, incomplete or full of errors, AI amplifies those problems. As she put it, nobody has perfect Salesforce (or really any CRM) data.

2. Give AI specific jobs, not vague ones

Vendavo did not roll out one giant AI project, it built tailored skills for specific tasks. The clearest example is sales prep, work that used to take an SDR 30 to 45 minutes, preparing for a call and researching the account, now takes about five.

Lindsay's approach is the same each time, define the task, make sure the right data flows in, and let the skill do the work. Vendavo applies it to chores that go stale fast, like updating competitive battle cards. The moment you finish one, it is out of date again.

3. Keep judgment human, and expect review work

Vendavo applies the same principle to marketing that it applies everywhere. Use AI to help understand, communicate and respond better, but don't outsource judgment to it. You can produce a lot of content, more content is not automatically progress, helping the right buyer make a better decision is.

There is also a cost to this speed. Lindsay pointed to a recent report from Anthropic, saying its employees now spend more than half their work week reviewing AI output. Whatever the exact number, the risk is real: AI can produce content and work faster than any person can absorb. The goal is results, not more activity.

The honest part of the conversation is worth hearing too. AI has made work move faster, but it has not freed up time for the strategy sessions people hoped for. The work continues, and the pace has accelerated, even small things take effort. Getting AI to stop adding filler sentences to a draft can feel like arguing with it more than it is doing the work.

4. Measure pipeline, not labels

Lindsay's scorecard is short. Are MQLs converting into pipeline, and are those deals closing? When Lauren asked whether AI should change how teams report on the MQL, Lindsay's answer was that relabeling it adds little. Her time is better spent generating more pipeline. If you can't see where pipeline comes from, that is the problem worth solving.

The same logic applies to AI projects. She shared that most proof-of-concept AI projects never reach production, because teams underestimate the engineering and maintenance cost and struggle to measure value. Start with the practical problem you are solving.

5. Know where humans stay in the loop

AI speeds up the research. Vendavo uses Semrush and Claude skills to find content gaps for AEO and SEO far faster than the old keyword-research grind. But strategy and positioning still need people. In Lindsay's words, AI does not replace her planning conversation with the Chief Sales Officer about pipeline and bookings.

It matters even more in a complex sale. Vendavo sells pricing software to very large manufacturers. A $10 billion company may research on Google and ChatGPT, and may chat with Vivi, but scope, cost and fit are conversations with several humans. AI helps by getting a sales rep up to speed on an unfamiliar industry. It does not replace the rep. Even when Claude drafts a campaign brief and the marketing platform builds the assets, a person reviews the output before it goes out.

6. Put the agent where buyers ask questions

During a brand refresh and website redesign, Vendavo made a deliberate choice. It did not tuck an agent into the corner of the page. Vivi, the Docket-powered agent, sits front and center in the hero. Buyers don't have to click around and work out their own journey. They can ask.

The reasoning is straightforward. If you talk to customers about AI, their first experience with you should show that they get good, accurate information right away.

Two things stood out after launch.

Buyers use more than English. About 15 to 20% of Vivi chats happen in another language, mainly German, Spanish, Portuguese and Japanese. Vendavo has not had to translate its site into each one.

The conversations are a content source. Sales calls tell you a lot, and agent conversations add what visitors ask before they ever talk to sales. For a marketing team, that is a steady stream of ideas for content.

None of this happened by accident. Vendavo spent a lot of time training and testing the agent and setting guardrails before launch.

Common questions

Where should an AI agent live on a B2B website? Vendavo put it in the hero so buyers can ask questions immediately instead of navigating the site.

Do we need to translate our site to serve global buyers? Not necessarily. About 15 to 20% of Vendavo's agent chats are in other languages, and the agent answers them without a translated site.

What should we do before launching an agent? Train it, test it and set guardrails. Vendavo invested in all three before going live.

Will AI save my marketing team time? It makes work faster, but it may not free up time. Plan for a quicker pace and more review, not an empty calendar.

The TLDR takeaway

Start with the buyer. Fix your data. Give AI specific jobs. Keep people in charge of judgment. Then put an agent where buyers already have questions, and read what they ask.

Watch the full conversation on demand: How an AI-First Company Builds Its Marketing Strategy Around AI.