I spent a day recently at Salesforce’s AIforce World Tour Essentials at ICC Sydney, and the idea I walked out with was a simple one. Agents take the admin. People keep the judgement.
It’s a line I’ve been testing in my own work for a while, so it was good to hear it from the stage.
What an agentic enterprise is
The session in the Agentic Enterprise Theatre started with a definition worth borrowing. An agentic enterprise is one where people, AI agents and platforms work as one team, and the work happens where people already are (Slack, Teams, Claude or a CRM) instead of in a dozen open tabs.
What I liked is that it starts with people. Salesforce framed the shift around 4 Rs: redesign, reskill, redeploy and rebalance. The technology is the easy part; deciding what it should do is the work.
Their step-by-step guide to becoming an agentic enterprise breaks it into 6 steps: align on the goal, design the human-AI workforce, start with one clear use case, redesign the process, ground agents in trusted data, and keep improving.
Start with the boring, repetitive work
The most useful slide was a set of 3 questions for choosing where an agent should go first:
- Which repetitive or high-impact tasks would benefit most?
- What’s your appetite for risk, and what data does the agent need?
- How will you measure impact, with both leading and lagging indicators?
That matches what I’ve seen building AI workflows for a marketing team. The quick wins were the repetitive jobs: sorting the shared inbox into tasks, preparing briefs, writing first drafts of case studies and emails. Someone still approved everything before it went out, and the time saved went back into the thinking.
The numbers on the day backed this up. According to Salesforce, 99% of organisations running agents keep humans overseeing them, and its own Slackbot rollout kept 80% of users while saving up to 20 hours a week. The honest limit is data: Salesforce’s guide also notes that most organisations say their agents struggle to reach current, accurate information. An agent is only as good as what it can see.
Try this: list 5 tasks your team repeats every week, then circle the one with the clearest rules and the cleanest data. That’s your first agent.
If you want the worksheets, Salesforce shares a use-case prioritisation sheet and a vision and system-mapping worksheet.
Where marketing is heading
The main-stage demos made it concrete for marketers. The one that stayed with me was Palmata, an answer engine optimisation tool from Contentful (now part of Salesforce). It shows how AI answer engines such as ChatGPT and Gemini describe your brand, and what to change so they describe it accurately.
It’s a useful reminder. More buyers now ask an AI assistant before they ever reach a website, so how those assistants talk about you may soon matter as much as where you rank in search. You can run a first Palmata report for free and see what they say about your brand today.
Learning out loud
My first time in Salesforce was way back at Herbalife, sending distributor emails in what’s now Marketing Cloud, so it was fun to see how far it has come. The best part of the day, though, was catching up with my dear ex-colleague Palisha, who now runs her own business, Apuki Life. Swapping notes between sessions is half the value of days like this.
The tools will keep changing. The question underneath them stays the same: which work should a person own, and which work can a machine take off their plate?