AI Won't Create Your Demand Gen Strategy, But It Will Make Execution 10x Faster

You’ll probably agree that Linkedin is full of AI “success” stories now. Someone's AI assistant "10x'd their pipeline" (when they probably hired three SDRs). Another “guru” promises their prompt will "revolutionise your ICP research" but what it'll give you is generic personas that sound like they came from a marketing textbook.

AI won't fix a broken demand gen strategy. It won't magically identify your ICP, write positioning that resonates, or turn weak offers into conversion machines. AI isn’t good enough (yet) to do that, but what it can do is make execution 10x faster once you know what you're building.

Unless you know what you’re building and who you’re building it for, AI can only help you so much. Here’s how AI can help your demand gen engine - particularly if you’re running a lean team and under pressure to deliver.

Why Most AI for Demand Generation Advice isn’t Helpful

You’re a marketing leader with tons of experience building strategies across different businesses and product lines over many years. You’ve executed campaigns that have turned the dial, you’ve been in high-pressure meetings and have had to explain to senior stakeholders why things haven’t worked out and what you’re going to do about it. You’ve also got a ton of domain knowledge about your business, product and sector. This experience can’t be substituted by AI. AI can’t replace strategic thinking - but it can accelerate execution.

You’ve already seen loads of thought leaders on Linkedin extolling the virtues of AI marketing tools for B2B and claim that it can do everything. You’ve seen them claiming that they’ve replaced their whole team with an agent workflow that got similar or better results at a fraction of the cost of their previous headcount. You've seen the AI hype machine claiming massive cost savings from replacing entire teams. Most have either massive headcount already or, like Klarna's well-documented AI experiment, terrible results that required an about-turn..

Most people now can spot who is using ChatGPT for demand generation content from a mile off. It’s not authentic, and to be honest, it’s a massive risk for those going all in on it as it’s eventually going to erode brand trust.

The real AI opportunity is with the stuff your customers can’t see. Marketing automation with AI can compress 6 hours of repetitive, high volume tasks into 45 minutes (or better). Imagine how much high-impact work you could achieve if you had an agent working in the background, doing all of the things that numb your brain that you hate doing?

The Three Questions Before You Use AI for Anything

7 Use Cases Where AI Can Actually Save Time

ICP Research & Validation

What: Using AI to review and analyse customer interview transcripts, support tickets, and sales call notes into patterns

Why it works: You're not asking AI to invent your ICP. You're using it to process 50 hours worth of work analysing qualitative data in 30 minutes and spot recurring pain points, objections, and language patterns.

Copy-and-paste prompt:

ROLE: You are an expert customer research analyst specializing in identifying pain points and buying triggers.

CONTEXT: I'm analyzing customer research for [product/service that does X for Y audience]. We've conducted interviews with customers who recently purchased. I need to identify patterns in their decision-making process to refine our ideal customer profile and messaging.

ACTION: Analyze these interview transcripts and extract:

1. The 3 most frequently mentioned problems customers had before finding our solution

2. Exact phrases and terminology they used to describe these problems (verbatim quotes)

3. Solutions they tried before us and specific reasons those solutions failed

4. The trigger event or "last straw" moment that made them actively search for a solution

5. Any contradictions or outliers in their stories

EXPECTATION: Format your analysis as a structured table with clear categories. Flag any contradictions between interviews. Highlight the top 3 most consistent patterns across all transcripts. Include quote examples for each major finding.

[Paste transcripts here]

Run this separately for your best customers vs. churned customers. You should find some gold nuggets. Remember though, AI can’t tell you which patterns matter strategically - that’s what you’re good at.

Landing Page Copy Variations

What: Generating 5-10 headline/subhead variations to test, based on proven frameworks

Why it works: Copy testing velocity is the difference between 2 experiments per quarter vs. 2 per month. AI makes it trivial to generate variations worth testing - you still validate via data.

Copy-and-paste prompt:

ROLE: You are a direct response copywriter specializing in landing pages with expertise in conversion-focused headline frameworks.

CONTEXT: I'm creating variations to A/B test for a [product category] landing page. Our target audience is [ICP with specific role/seniority]. Our current headline is: "[current headline]". Our core value proposition is: [one sentence]. The main objection preventing conversions is: [specific barrier].

ACTION: Generate 8 distinct headline variations using these proven frameworks:

Problem-solution format (lead with customer pain)

Outcome-focused format (lead with specific result)

Social proof angle (lead with credibility/customer success)

Urgency/scarcity angle (lead with time-sensitivity)

Question format (provoke self-reflection)

Contrarian angle (challenge common assumptions)

EXPECTATION: Each headline must be:

Maximum 60 characters

Written in conversational B2B tone (not corporate jargon)

Clearly different from the others in approach

Paired with a one-sentence explanation of the psychological trigger it's testing

Present as a numbered list with the framework name, headline, and rationale for each.

UTM Taxonomy & Campaign Naming

What: Building consistent, scalable UTM structures and campaign naming conventions

Why it works: It's tedious, rule-based work that AI handles perfectly. One prompt builds a 6-month taxonomy in 3 minutes.

Copy-and-paste prompt:

ROLE: You are a marketing operations specialist who builds scalable UTM parameter structures and campaign naming conventions.

CONTEXT: We're launching demand generation campaigns across multiple channels and need a consistent UTM taxonomy to track attribution accurately. We run these campaign types: Paid Social, Paid Search, Email, Webinars, Content Syndication. Our channels include: LinkedIn, Google Ads, Newsletter, Partner sites. Our target ICPs are: Series A SaaS companies, Series B SaaS companies, and late-stage startups.

ACTION: Create a comprehensive UTM parameter structure that includes:

- utm_source (standardised source names)

- utm_medium (channel categorization)

- utm_campaign (format: YYYY-MM_campaigntype_audience_offer)

- utm_content (for ad creative and A/B test variation tracking)

- utm_term (for paid search keyword tracking where applicable)

EXPECTATION: Deliver:

1. A clear taxonomy table showing all parameter options

2. Naming convention rules written as if for a team wiki

3. 10 complete URL examples showing the structure in practice across different campaign types

4. A "common mistakes to avoid" section

Format as a reference document that a junior marketer could follow without asking questions.

Export this to a Google Sheet your team can reference. Consistent UTMs = cleaner attribution data.

You still need to define why you're tracking what you're tracking. AI can't decide your measurement strategy.

Competitor Ad Creative Analysis

What: Analysing competitor LinkedIn/Google ads to identify messaging patterns and creative angles

Why it works: Competitive research is time-consuming. AI can process 50 competitor ads in minutes and spot patterns you'd miss manually.

Copy-and-paste prompt:

ROLE: You are a competitive intelligence analyst specializing in [your industry] advertising and messaging strategy.

CONTEXT: I've collected 20 LinkedIn ads from our direct competitors in the [industry/category] space. We sell [product/service] to [ICP]. I need to understand competitor messaging patterns to identify positioning gaps and creative opportunities we can exploit.

ACTION: Analyze these competitor ads and create:

1. Common headline formulas they use (with 2-3 examples of each pattern)

2. Value proposition angles they emphasize (outcome, speed, cost savings, ease of use, etc.)

3. Call-to-action patterns (free trial, demo, download, assessment, etc.)

4. Visual creative patterns (product screenshots, team photos, data visualization, illustrations, etc.)

5. Messaging weaknesses or gaps in their positioning (problems they're not addressing, objections they're not handling)

6. Language patterns (formal vs casual, feature-focused vs benefit-focused, technical vs accessible)

EXPECTATION: Format as an actionable creative brief with:

- Pattern summaries backed by specific examples

- A "white space opportunities" section highlighting what competitors are NOT doing

- 3 specific recommendations for how we can differentiate our creative

- No generic observations. Every insight must be specific and actionable

[Paste ad copy and describe visual elements here]

Pro tip: Use Meta Ad Library and LinkedIn Ad Library to collect competitor ads systematically. You still need to put in some manual effort here, but AI can take hours away from analysis.

Warning: Don't copy competitors. Use this to identify white space in messaging positioning and create an advantage.

Email Subject Line Testing

What: Generating 10-15 subject line variations optimised for B2B open rates

Why it works: Subject lines are high-volume, low-stakes testing opportunities. More variations = faster learning.

Copy-and-paste prompt:

Generate 12 email subject lines for [campaign objective].

ROLE: You are a email marketing specialist focused on optimizing subject lines for professional audiences with high inbox competition.

CONTEXT: I'm creating an email campaign for [specific campaign objective, e.g., "inviting prospects to a webinar on demand generation challenges"]. The audience is [ICP role + seniority level, e.g., "Heads of Marketing at Series A SaaS companies"]. The key message we want to communicate is: [one sentence]. This email is part of [campaign type: cold outreach/nurture sequence/event promotion].

ACTION: Generate 12 subject line variations that test different psychological approaches:

- Curiosity gap (tease the value without revealing everything)

- Direct benefit (state the clear outcome they'll get)

- Question format (provoke self-assessment or reflection)

- Urgency (time-sensitive angle without feeling spammy)

- Personalization (role-specific or industry-specific relevance)

- Contrarian (challenge a common assumption in their field)

- Social proof (reference customer success or adoption)

- Data/stat hook (lead with an interesting number)

EXPECTATION: Each subject line must:

- Stay under 50 characters (mobile-friendly)

- Avoid spam trigger words ("free," "urgent," "limited time," "act now")

- Be appropriate for B2B professional audiences (no clickbait)

- Be paired with a one-sentence explanation of which psychological trigger it's testing

Format as a numbered list grouped by approach type, with a brief note on when each approach works best.

Pro tip: A/B test 2-3 subject lines per send. AI accelerates hypothesis generation, not validation. Don’t overdilute your tests. More variations mean you need a greater sample for statistical significance.

Warning: Most B2B audiences have subconscious clickbait filters. Prioritise clarity over anything too clever.

First-Draft Ad Copy (Paid Social)

What: Creating 5-8 ad copy variations for LinkedIn/Facebook testing

Why it works: Paid social needs constant creative refreshment (ad fatigue hits very quickly). AI makes it feasible to test new copy variations more regularly.

Copy-and-paste prompt:

ROLE: You are a paid social advertising copywriter specialising in LinkedIn and Facebook campaigns that drive qualified lead generation.

CONTEXT: I'm creating ad variations for [product/service] targeting [detailed ICP with pain point]. The campaign objective is [awareness/consideration/conversion]. Our key message is: [value proposition in one sentence]. The call-to-action will be: [demo/download/free trial/assessment]. These ads will run on LinkedIn and need to stop scrolling professionals who see 50+ ads per day.

ACTION: Write 6 LinkedIn ad variations (primary text only, not headlines) that test different messaging angles:

- Problem-agitate-solution format (describe pain, amplify it, present solution)

- Customer story angle (brief success narrative from real user perspective)

- Stat/data hook (lead with compelling number or benchmark)

- Question opening (provocative question that forces self-reflection)

- Contrarian positioning (challenge conventional wisdom in the space)

- Before/after transformation (contrast old way vs new way)

EXPECTATION: Each ad variation must:

- Be maximum 150 characters (LinkedIn truncation-friendly)

- Use conversational B2B tone (avoid corporate jargon and buzzwords)

- Include a natural transition to the CTA

- Have a clear single message (not trying to say everything)

- Be distinct enough that performance differences will be meaningful

Format as numbered variations with a note on which audience segment or funnel stage each might perform best for.

Pro tip: Pair this with real customer language from sales calls. Generic AI copy gets ignored.

Warning: Always review for brand voice. AI defaults to bland professional-speak. You need edge and personality. Adding your brand tone of voice guidelines into the prompt (or a file in the project) can help with this.

Meeting Prep & Briefing Docs

What: Synthesising prospect research (website, LinkedIn, news) into a pre-call brief

Why it works: Sales reps spend 30-45 minutes researching each prospect. AI does this in 5 minutes while maintaining quality.

Copy-and-paste prompt:

ROLE: You are a sales intelligence researcher who creates pre-call briefing documents for B2B sales teams to help them have more informed, relevant discovery conversations.

CONTEXT: I have a discovery call scheduled with [prospect company name]. They are a [company description: size, industry, business model]. I need a concise briefing document that helps me understand their business, anticipate their challenges, and ask intelligent questions. We sell [your product/service] which helps companies [core value proposition].

ACTION: Research and synthesise information from these sources to create a discovery call brief:

1. Company overview: What they do, who they serve, business model, key products/services

2. Recent developments: Funding rounds, product launches, leadership changes, partnerships (last 6 months)

3. Technology stack indicators: Tools mentioned in job postings, integrations on their website, tech visible in their product

4. Likely pain points: Based on their industry, growth stage, and business model, what problems are they probably facing that we can solve?

5. Strategic context: Market position, competitive pressures, growth trajectory

6. Intelligent questions: 3 specific, thoughtful questions to ask in discovery that demonstrate we've done our homework

EXPECTATION: Format as a scannable one-page brief with:

- Bullet points and short paragraphs (easy to skim in 3 minutes)

- Clear section headers

- Specific details (not generic observations)

- Sources cited for key facts (so I can reference them if needed)

- "Red flags or qualification concerns" section if anything suggests they might not be a good fit

[Paste relevant URLs, LinkedIn profiles, recent news, job postings, or company website content here]

Pro tip: Use this for ABM research at scale. Personalise outreach without hiring researchers.

Warning: AI can't replace genuine curiosity in discovery calls. Use briefs as a starting point, not a script.

The 4 Things AI Can't (And Shouldn't) Do for Your Demand Gen

Strategic positioning

I mentioned this earlier in the article. LLMs are based on texts and content that has already been published and often has a knowledge cutoff. It’s more than likely going to regurgitate what’s been said 1000 times already (in different ways) and could well be out of date. It doesn’t have your experience or intuition, and you understand your market better than its training data.

Nuanced Buyer Psychology

AI doesn’t understand unstated objections, office/project politics or tensions within a buying committee. It can tell you what your customers say, but it won’t tell you what your customers actually mean when they say “we need to speak about this internally”. You’ll probably get a gut feeling about this - AI can’t replace your intuition.

Channel Strategy & Budget Allocation

AI doesn’t know what your current CAC payback constraints are, what your growth stage priorities are, or about those conversations you’ve had with finance about Marketing’s ROI. If you ask ChatGPT how to allocate a £50k budget, it’ll give you a generic split. Anyway, your channel strategy should be predicated by your campaign goals and should be quite fluid, in this day and age.

Creative Strategy & Brand Voice

If you pop over to your Linkedin feed now, you’ll probably be able to spot a post written by an AI easily (emojis for bullet points, overuse of em dashes, superficial analysis, that kind of thing). They’re trained on pre-existing content, so it’s more than likely going to produce content that’s safe. This tends to excite and/or offend nobody, which in turn isn’t going to command the attention of your target audience or followers.

Don’t rely on AI to replace your deep thinking around the above points. Instead, treat AI as a junior member of your team. It needs direction, but is blazing-fast at churning out the repetitive tasks that you don’t like.

How to roll-out AI in lean teams

1. Start Small

AI is great at automating tasks at scale, which is excellent at saving time. Google claims users can save 133 hours with just 3 hours of AI training. Sounds great, but this also means that unsupervised, AI can create havoc and cause irreparable damage if you just let it do it’s thing without guardrails.

To prove its value, pick one small use case (such as the ones above), and build an MVP. Run a few prompts, review the output, and then refine your prompts (if needed). Benchmark the time it takes you to do this vs. doing it manually - this should help your business case if you’re building one, or at least confirm (or deny) that AI is helpful in the context of the task.

2. Build your prompt library

Document the prompts that have worked in step one. This is important for scaling your AI output as you can then refer back to it if it goes wrong at any point. Also, when new models are released your prompts may not give you the same output.

Share the prompt library with your team in a central, version controlled document. Then you can scale across your team and save more time. Make sure that if you make refinements, that you document them too. Even a small note that says “this works better when you add x” gives you some history on the prompts, which you can then potentially use on future prompts and projects.

3. Scale

Try a couple of more use cases (e.g. ad copy or email subject line variations) and repeat step 1. The time savings should compound. If you’re happy with the output (and have the budget) you can start building agents for these tasks (e.g. N8N, ChatGPT’s native Agent, or one of the many other workflow automation tools on the market), that will automate the process for you. Remember the output is only as good as the prompt you feed the AI, and always review the output before publication or putting into production. Blind faith in the AI could cause irreparable damage to your brand or trust.

4. Measure and Iterate

Review all of your outputs from the month, and identify where AI helped and where it hasn’t. Adjust your prompt frameworks based on your learnings. Review how much time you have potentially saved in the month using AI, and think about other areas in your team where you can apply what you’ve learned.

The key thing here is learning to crawl before you can walk and run. Make sure that AI helps with your smaller tasks properly and learn from the outputs and mistakes. This will help you no end in scaling. One use case done well beats 3 or 4 implemented poorly.

AI is a multiplier, not a replacement

If your demand gen strategy isn’t optimal, AI will just help you fail faster at scale. An unclear ICP, weak positioning, no measurement framework will just multiply the issues you’ll see.

But if you know what good looks like? AI becomes the junior team member who works weekends, never complains, and handles the high-volume grunt work that slows you down and exacerbates your frustrations.

The competitive advantage isn't simply just using AI. Everyone has access to ChatGPT. The advantage is knowing:

• Which tasks to delegate vs. which to do yourself
• How to prompt for quality, not just speed
• When to trust the output vs. when to bin it

Lean teams that master AI utilisation will ship more experiments, test more creative variations, and move faster than competitors with twice the headcount.

The question isn't "Should we use AI?" It's "What are we still doing manually that AI could compress from 6 hours to 45 minutes?"

Start with one use case this week. You'll know within a few days if it's working.

Photo credit: Shantanu Kumar: https://www.pexels.com/photo/get-instant-answers-to-your-questions-with-chatgpt-your-pocket-ai-assistant-stay-connected-with-chatgpt-on-the-go-16689016/