A campaign strategy in hours, still in my own words

The challenge

Every campaign started with a strategy document, and each one needed the same pieces: the goal and KPIs, the target persona and audience, the buyer’s journey stage, the channels, budget, and timeline, benchmarks to beat, our past results, and what was happening in the industry. Most of that already existed somewhere. It was spread across persona documents, brand and tone of voice guides, past reports, and my own notes. Pulling it together, researching benchmarks, and writing it up from scratch each time took more time than it needed to.

I also wanted the strategy to be checkable. When I took a plan to leadership, I needed to show why each decision was made, with sources for every benchmark and best practice.

I didn’t want to ask AI for a strategy either. Give it a goal and a channel, and it fills the gaps with what it already knows, so you get a rehash of every other campaign plan online. A good strategy needed every element in one place: my goals and thinking, our personas and USPs, our brand and tone of voice, our past results, benchmarks, industry news, and what competitors were doing. Doing that properly by hand meant checking one document against another, and it was easy to miss something. Leave any of it out, and the plan is less informed than it could be. I wanted the AI to bring all of it together around my own thinking, so nothing was missed and the decisions stayed mine.

My approach

Give the AI our context once. I set up a project in ChatGPT and added our core documents to its project knowledge, which the AI could refer back to in every chat: our target personas, USPs, values, brand guidelines, tone of voice guide, and reports on past campaign results.

Talk, don’t type. For each new campaign, I dictated my thinking into Word, covering what we wanted to achieve, who it was for, when, and how. Speaking let me get my scattered ideas down far faster than typing them. Stanford researchers found speech recognition was about three times faster than typing on a phone keyboard.

The AI sorts, I decide. The AI’s job was to put my words into the right sections of the template, keep my wording, and fill in what it could find in the project knowledge. I made every decision: the goal, the timeline, the budget, the persona, the audience, and the channels.

Source: Stanford University, “Smartphone speech recognition can write text messages three times faster than human typing” (Ruan, Wobbrock, Liou, Ng, and Landay).

What I built

A strategy template. I used AI to build a standard structure from our past campaign plans, so every campaign followed the same format. It covered:

• The basics: campaign name, goal, KPIs, timeline, budget, channels, and who was involved
• Audience: the target persona, the specific segment or account list, its size, and where they were in the buyer’s journey
• Benchmarks: industry benchmarks for each channel and KPI, and our own past results
• Research: industry news and trends, competitor activity, and channel best practices
• Measurement: what we’d report on, how often, and who to
• Recommendations: what to test, and what we could change or expand

Personas on hand. I only had to name the persona, and the AI pulled the details from our persona document into the strategy. For the audience size, I built the segment in HubSpot or a company list in Firmable and added the count myself.

Sourced benchmarks. For each channel in the plan, I asked the AI to find benchmarks for our industry in Australia, such as email open rates, Google Ads click-through rate, and cost per lead, and to link every one to its source so I could check it. It also added our own past results for each channel from our reports, so we knew both the industry figure and our own figure to beat.

Industry research. I had a set of prompts, written with ChatGPT, for researching what was happening in our target market in Australia: news, regulation, and the problems buyers were facing. For an aged care campaign, that meant looking at new regulations and accessibility requirements for websites, such as WCAG (the Web Content Accessibility Guidelines). I then looked for where those issues overlapped with what we could help with.

Best practices and competitors. I had the AI research best practices for each channel in the plan, with sources, so leadership could see why each decision was made. For a Google Ads campaign, I also collected the competitor ads showing for our target keywords and had the AI review them against those best practices and our USPs, to see where our message could stand apart. This built on our wider competitor analysis.

A gap check at the end. Once the draft was filled in, I asked the AI what was missing and what it would recommend, based on the research, best practices, past results, and our goals.

Checking everything. I read every section before using it. I checked the name, timeline, persona, and segment against what I’d said, and checked benchmarks against their sources, especially anything going to leadership.

Results

• A few hours instead of a day. A strategy that used to take me about a day to research and write took a few hours of working with the AI.
• More time on the campaign. The time I saved went into building, testing, and learning from the campaign, instead of planning it.
• The same structure every time. Every campaign used the same template, so nothing was skipped and anyone could find their way around a plan.
• Plans leadership could check. Benchmarks, best practices, and past results were linked to their sources, so each decision had a reason behind it.
• My thinking, my words. The strategy came from my brief and our own information, with the AI collating and checking it.

What I’d do next

Today I’d build this in Claude, which I now use for most of my work. A Claude project holds the same documents as project knowledge, with standing instructions to keep my wording and link every source.

I’d add the original dictation to the bottom of every strategy document. If I later questioned why the AI had included something, I could check exactly what I’d said.

I’d also add a forecast section. With the budget, channel split, and our past cost per lead for each channel in the project, the AI could estimate the leads we might expect, and I could check that the timeline and budget were realistic for the goal.