Boost your sales outreach with AI

Most consultancy outreach fails for boring reasons. These include the targeting being a bit vague, the message being generic, the follow-up being inconsistent, and the whole thing competing with a prospect’s already overloaded inbox.

Umbar Shakir, former partner at Gartner and Chief AI Office at Havas, delivered a workshop to our members on using gen AI for sales outreach. The point she kept coming back to was that outreach is not a dark art, it is a system. You will already know the system, you have your funnel maths. As she put it, you will also have ‘rubrics’ for the volume you need: “how many people do I need to be reaching out to, to get those meetings?”

According to Umbar, Gen AI doesn’t change the need for that discipline. It changes the friction within it. AI will help you do the prep faster, personalise without losing your mind and follow up with more consistency. Used well, it helps you be sharper in every interaction. The flipside is that used badly, it just makes you noisier.

Based on Umbar’s workshop, here’s a practical way to use gen AI to lift outreach quality without turning it into spam.

1. Use AI to get specific on who you are targeting

Outreach is not a dark art, it is a system

Outreach gets easier when your ‘who’ is really tight.

Gen AI can help you build mini ‘target briefs’ in a really useful way, such as what this role is likely measured on, what would make them say ‘yes’ to a first conversation and what might be a credible trigger for change right now.

The key, as with all AI use, is to treat the output as a starting point for your thinking, not a replacement for it. The fastest route to mediocre outreach is asking AI to guess your ICP and then sending whatever comes back. Instead, Umbar recommends using it to generate your hypotheses, then sanity-check them against what you already know about your market.

According to Umbar, a simple internal habit that helps is to keep a running list of the top 10 problems your best prospects talk about, then feed that into your AI briefs so that it’s anchored in reality.

2. Turn research into a hook, not a data dump

Most ‘personalised’ outreach is just a fact about the prospect’s company, followed by a pitch.

Umbar explains that that’s not a hook. It’s lazy, and you’ll know when you see it in your own inbox that you can see right through it.

Umbar says that a good hook is a point of view: a credible, specific reason why a conversation is worth their time. “Gen AI is useful here because it can help you explore angles quickly” says Umbar. “Give it your sector context and ask it to propose 3–5 hypotheses for what might be driving change for that buyer right now, then pick the one you believe will prompt a response.”

This is where consultancies have an advantage. You have patterns from client work, and AI can help you package those patterns into sharper language, but it can’t replace your judgement.

3. Draft outreach that sounds like you, not anyone else

Used well, AI helps you be sharper in every interaction. Used badly, it just makes you noisier.

Umbar emphasised the point of being specific. Instead of using a prompt like ‘draft an outreach email to get a meeting’, she shared a far more precise and effective instruction. Members can watch the full workshop on the Growth Hub. Not yet a member?

If you want gen AI to produce something usable, you need to give it real inputs: the prospect’s role, what you believe matters to them, your offer, your proof points and the tone you actually use. Otherwise you’ll end up with a polished version of nothing.

One warning from Umbar that is especially relevant to outreach: “You might write an email or a proposal using AI, hit send and your brain forgets it instantly because it wasn’t your thinking”. If you outsource the thinking, you also outsource conviction. Prospects feel that.

Umbar shared a particularly practical rule: “If you wouldn’t be comfortable saying the message out loud in a sales call, it’s not ready to send.”

4. Build follow-up discipline without spending your life writing emails

Most teams either under-follow-up (‘I don’t want to pester’) or over-follow-up with low-value nudges (‘Just bumping this’). Neither works.

Gen AI is genuinely useful for follow-up because it can turn notes into a coherent, structured message quickly, as long as you provide the substance. Umbar described a simple approach: “If you’ve got a transcript of a meeting, you can upload it and prompt AI to write you a follow up email: contain the greeting, the recap, the resources and the next steps”.

This doesn’t replace building relationships, but it does ensure the basics happen clearly and consistently, every time.

If you outsource the thinking, you also outsource conviction. Prospects feel that.

Even without transcripts, you can apply the same idea. For example, feed the model your bullet notes and ask it to draft a follow-up with:

• a two-line recap of what they said mattered
• the one most useful resource you promised
• a clear next step with a date or decision.

As Umbar says, you still own the judgement, AI just helps you get it on the page faster and well-structured.

5. Let your CRM do some of the heavy lifting

If you want outreach to improve, you need feedback loops. That means using your CRM properly, not just as a contact graveyard.

Umbar’s tooling point was clear: for prospecting and outreach, look for “the ability for a tool to integrate with your CRM system or your LinkedIn.” Then “start using your CRMs properly if you want to really leverage AI”.

Why? Because AI gets better when it can learn from your history. Which subject lines get replies, which hooks land by sector, what follow-up cadence works and what objections repeatedly came up.

A simple pilot to run this week

If you want to test this without boiling the ocean, we recommend running a two-week outreach pilot with one niche and one offer.

• Choose 20 of your warmest prospects
• Build a one-page ‘target brief’ template and use AI to create a first draft for each prospect
• Write one core message, then create two variants (different hook, same CTA)
• Send, follow up twice, then review replies and patterns before scaling.

The goal isn’t to send more outreach, but to turn your existing outreach into something that sounds more informed, more relevant and more deliberate. Used this way, gen AI becomes a sales multiplier, not a spam generator.