Improving AI marketing output with better briefs, context, and prompting

Went through the Spotlight announcements this morning. Writing up what seems to matter and what I think is oversold, in case it saves someone the time

Context on me: I work on HubSpot portals for a living, based in Australia, mostly 50 to 500 staff B2B.

The theme this year is "growth context". The pitch is that your data records what happened and context explains why, and that most portals have the first and none of the second.

What I think genuinely matters:

Context Home. The most boring announcement and probably the most consequential. Most portals carry an auto generated description of the business that nobody has read or corrected, and every AI output in the portal is built on top of it. If yours is wrong or three years stale, that is why your AI output has been disappointing. Read yours before you touch anything else.

Agent Builder triggers. Workflows can now start from Slack, a Google Sheet, a webhook or a schedule. If you pay for Make or n8n mainly to get scheduled triggers, that subscription is worth auditing now. There is also a one click migration from a legacy workflow to an agentic one, which I am sceptical of until I have broken one.

Marketing Studio 2.0. HubSpot said openly that the 2024 version was hard to navigate and that its AI ignored brand context. It now reads brand identity, ICP and past campaign performance from the portal. Which means output quality is set by how well those fields are filled in, and in most portals they are not.

What I would not rush: HubSpot Work. AI first workspace, projects and docs and tables on live CRM data, aimed at ClickUp and Asana and Notion. Private beta, and it will be paid. Interesting, but moving a team's project management onto a private beta is a bad trade.

The nurture agent. It rewrites subject line, body and call to action per person at send time, on top of your existing workflows. Powerful, and it means you can no longer see what a given person received by opening the email. If you run support, sort that out first. Unclear to me: Pricing on most of it. Whether the agents are gated by tier in a way that hurts smaller portals.

APAC availability and timing, which historically runs well behind the keynote. If you are in Australia or New Zealand, do not plan a quarter around a date announced in Boston.

The Breeze name is also being retired. Breeze Assistant keeps it, everything else drops the prefix.

Mostly cosmetic, but worth a pass over any internal documentation you wrote in the last two years. One Australian specific: agents that read and rewrite customer communications sit inside the Privacy Act changes due 10 December, so that is a conversation worth having early rather than late. We costed the existing agents in Australian dollars a while back and the method still applies to the new ones, happy to link if useful. Disclosure, that is our own site.

Anyone had hands on any of this yet? Particularly interested in whether the workflow migration is real.

Things I wish I knew about AI when I started using it for marketing #1

I thought learning AI meant learning tools.

It took me a while to realise that wasn’t the hard part.

The hard part was learning how to think with AI.

When I first started using AI for marketing, I would give it a prompt, get an answer, and think:

Okay… this is useful.

But the more I used it, the more I realised something:

The quality of the output is often a reflection of the quality of the thinking behind the prompt.

If I ask:

Give me 10 campaign ideas for this brand.

I’ll get 10 campaign ideas.

But if I first give it:

→ the brand positioning
→ the target audience
→ the campaign objective
→ what we’ve already tried
→ the consumer insight
→ the tone of voice
→ the constraints

the conversation completely changes.

Now I’m not asking AI to think instead of me.

I’m giving it enough context to think with me.

That’s when AI became genuinely useful in my workflow.

Today, I use AI to:

• get through the first draft faster
• challenge my ideas
• explore different angles
• organise messy thoughts
• research and synthesise information
• automate repetitive parts of my workflow

But I still make the decisions.

Because AI can give me 20 ideas.

My job is to know which 2 are actually worth pursuing.

That’s probably the biggest lesson I’ve learned so far:

Don’t focus on becoming better at using AI.

Focus on becoming better at thinking and then use AI to multiply that thinking.

#AI #Marketing #MarketingStrategy #AITools #FutureOfWork

My manager sent the email on a Thursday afternoon.

“We need a marketing campaign by the end of the month. You’re leading it.”

No agency. No budget. No marketer on the team.

Only me. I stared at his message for a long time.

Then I did what every IT person does when they don’t know something — I started Googling. One tab turned into twenty. And that’s how I stumbled into the world of artificial intelligence in digital marketing.

I had no idea what I was getting into.

Here’s what actually happened.

𝗛𝗼𝘄 𝗱𝗼 𝗜 𝘂𝘀𝗲 𝗔𝗜 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝗹𝗼𝘀𝗶𝗻𝗴 𝘁𝗵𝗲 𝗵𝘂𝗺𝗮𝗻 𝘁𝗼𝘂𝗰𝗵?

This was one of the questions for today’s Founder Coffee: AI x Growth.

Let’s start here:

AI doesn't sound robotic because it is AI.
It sounds robotic because it lacks human direction.

AI-generated content can be spotted quickly.
Most often, it comes down to two key reasons: not having a strong sense of what good writing sounds like and not being clear on your own voice.

Three things you can do today:

𝟭. 𝗗𝗲𝗳𝗶𝗻𝗲 𝘆𝗼𝘂𝗿 𝘃𝗼𝗶𝗰𝗲 𝗯𝗲𝗳𝗼𝗿𝗲 𝘆𝗼𝘂 𝗱𝗲𝗹𝗲𝗴𝗮𝘁𝗲 𝗶𝘁
If you want AI to sound human, start with your own tone of voice.
How do you actually sound with clients, partners, or your community?
Clear? Warm? Direct? Strategic? Thoughtful?
AI can't reflect a voice you haven't defined.

𝟮. 𝗚𝗲𝘁 𝗯𝗲𝘁𝘁𝗲𝗿 𝗮𝘁 𝗴𝗶𝘃𝗶𝗻𝗴 𝗱𝗶𝗿𝗲𝗰𝘁𝗶𝗼𝗻𝘀
In addition to sharing strong examples, explain exactly what you want, step by step.
You’d be surprised how often AI misses the mark not because it's incapable, but because the direction wasn't specific enough.
Just like with people, the quality of the result depends on how well you communicate.

𝟯. 𝗞𝗲𝗲𝗽 𝗵𝘂𝗺𝗮𝗻 𝗷𝘂𝗱𝗴𝗺𝗲𝗻𝘁 𝗶𝗻 𝘁𝗵𝗲 𝗹𝗼𝗼𝗽
AI can produce a strong draft.
But you are the one who decides what should go out and what's missing.
That's one of the reasons why, in AIMAR, we keep a human layer for approval and process management.

So maybe the better question is not how to use AI without losing the human touch, but how to lead AI to amplify it.