As a marketing leader, how do you use today’s AI platforms and tools to become more efficient? Guillaume Cabane, Co-Founder and General Partner at HyperGrowth Partners and ex-head of marketing at Drift, Gorgias and Segment, shares how you can win with AI in outbound, SEO, and paid.
Deep Dive: AI-Powered Marketing to Get More Leads and Customers with HyperGrowth Partners
AI was good a year or two ago, but its quality was less than that of a human. There was less cost and less quality. Nowadays, we have more quality and speed, yet still at a lower cost. What does that mean for marketing?
The Real Impact of AI Today
Each motion has different CACs. On the far right, you see the sales-led approach we’re used to, which is pretty efficient but relatively expensive. On the left, you have PLG, which is fairly cheap but generally drives lower CAC and smaller customers.
You have this spectrum from low contract value to high ACV, which is correlated with low CAC to high CAC.
The ACV/CAC Spectrum
With those CACs and ACVs, you have typical segmentation of micro to Enterprise, and within that segmentation, you have the channels. On the cheap, low-CAC side, you’ll have SEO and virality: the PLG playbooks.
In the middle, you have outbound; at the far end, you have ABM and partnerships. That was yesterday. It’s what we used to do for the past decade.
Logical Conclusion of the AI Shift
What’s changing? The fact is that we’re now capable of moving faster with better quality. For example, we can scrape websites and random content for extremely cheap. We then feed that back into the AI prompt. For outbound and SEO, that drives an extremely low CAC.
What used to be mid-market channels can now be SMB channels because CAC is dropping. You can have new audiences.
You Can Get Mid-Market Prospects with SEO and Outbound
This distribution means you can acquire startups through outbound or Mid-Market through SEO. How does that work? You can reach Mid-Market prospects through SEO because you have better-quality content.
You create extremely high volumes of high-quality content when you have diverse content. On outbound, it’s the same thing. You’re no longer employing SDRs, which are a high cost.
That’s where we’ll hover around: the CAC-to-channel fit and how you have to adapt to this new reality.
Channel #1: Outbound
If you think back to 2015, what were the best SDRs doing?
- Information gathering and research
- Email content composing
- Outreach on LinkedIn
- Being persistent
This worked from 2015 to 2018. It was slow and expensive. In the middle, everyone started automating those sequences on Apollo or Apple markets, generating huge volumes at a low quality. With that, the conversion rate dropped.
Over the last year, we’ve become capable of scraping data at scale and generating high-value emails. This drives the conversion rates back up and is better than humans at a lower cost. This is the shift we’re seeing. If you look at this image, you will see a ton of data sources from which we can get great data.
Which SDR will go through at least one of the logos per data source to understand the prospect or customer and compose a great sequence? No one. They have minutes per email. But good AI can do that.
Example: Hands-on with Clay for Reddit
G’s latest advisory client, Reddit, gave him a goal to get 10% of U.S. advertisers to advertise on Reddit by the end of the year, starting from pretty much zero. It was a big goal.
The first idea was to scrape Facebook, LinkedIn, and Google Ad libraries, make sense of the ad, understand the company, aggregate that third-party data, and inject it into an email.
They went with Clay to prove the concept’s feasibility. The prompt explains how to figure out the Facebook page, find the ads, and, behind the scenes, make sense of the data.
For the email, they injected an ad image, explained the ad and company, and recommended the right subReddit to advertise on. The email is super personalized and makes a ton of sense.
You can build something like this automatically, and it works. Once it’s done, scaling it is extremely cheap. They did this in less than two months and beat the SDR team of 200 people who had been working on it for a couple of years.
“You can question the future of the SDR roles when you see things like that,” G says.
The Key Learning for Outbound
Sequences are dead. If you think the best way to engage customers is to have a series of pre-written, templated emails that you fire until someone responds, you’re living in the wrong era.
Every new email should be purposefully based on the situation, your data, and the existing conversation. And, all of that should be done automatically.
Channel #2: SEO
We’re coming off a world of pretty bad SEO. Many outsource SEO content creation to low-quality freelance writers to get on the search engine. This is primarily low-cost generation, low-value content, or tokenized and spinning content.
No one wants to read that. It worked until it didn’t. The playbooks of previous eras that did work were Zapier and G2. They have amazing PLG and user-generated content in B2B, which is very rare and hard to pull off.
G2 has a treasure trove of SaaS reviews that no one else has, and Zapier hired a team to build an integration catalog. Both cases can be done, but it’s very expensive.
Now, we can use LLMs for programmatic page creation and get the data without the workload. Not everybody can have a strong UGC motion, but everyone can be Zapier nowadays. How?
Let’s look at Deepgram.
Example: Deepgram
What they did is very typical SEO keyword research. Using Airtable to store the information, they have all these keywords and data. Using Airops, an LLM orchestration platform, they can have different sub-agents look for the best title, page, or screenshot and bring it back together into one prompt with the data injection.
That is a much better way to create content than having a huge prompt that tries to do everything. Try breaking it down, storing it in Airtable, and then feeding the data from Airtable to the final agent.
It’s a reasonably advanced workflow, but check out the results. Before 2023, people were trying to do this manually with low budgets. By automating it, they got 3.5M clicks. That’s wild!
They had one person to build a workflow, which took about 1000 hours. One human does ongoing control and validation, replacing a ten-person team. For now, using LLMs to get organic traffic from SEO is possible. Will it last? We’ll find out.
There’s a time window to explore this right now.
- Scraping websites
- Building unique content
- Enriching it with third-party sources
- Doing the quality control
The AI workflows produce better content than autonomous agents and better content than humans alone unless you’re willing to pay $10,000 per post.
Channel #3: Paid
Rex, the world’s best guy at paid, at Hubspot, who’s spent over $100M on Google search alone over the last five years, shares his take. Before, we had a fairly standard situation.
Most companies had an in-house paid team, an agency, or freelancers. They segmented ad campaigns by various customers and traits, brought them into ad groups with targeting and copy, and constantly fine-tuned bids.
Google has tried introducing AI support tools over the years, but they haven’t worked. That’s changed within the last six months. Now, they work and make the job seamless.
LLMs inside Google ads are improving the targeting, showing the right creative to the right audience, and setting the right bids. The conversions are higher than you could manually achieve if you tried to override the AI.
The problems solved are audience targeting and bidding. What remains for humans for now is the research part and the creative process. But wait! The creative process. Now, you can use AI to build and design ads based on your website. It will create video ads for you for $0.
What This Means For Paid Teams
That’s just the beginning. If you’re still trying to fight AI by segmenting campaigns into a ton of different budgets, you’re fighting the wrong war. Stop it. Let the ad network figure it out for you.
The more you handcuff the system and restrict the AI, the fewer optimizations it can make on your behalf, and the worse off you will be. This means that paid teams have jobs similar to those in SEO: smaller, more technical teams.
Your job is to build the infrastructure enabling AI platforms to learn. Data quality is critical, and it has to be able to learn the good and bad outcomes. Plus, you need the budget to ensure the volume of conversion so that it can learn.
Key Takeaways
- Across the board, generating leads involves fewer humans and more AI models. This is the final takeover of SDR into technical marketing roles.
- More than ever, this is a battle of CAC. The AI change is shifting the CAC of many channels and opening channels to marketers they hadn’t considered for those audiences.
- If you invest in technical people who can build the workflows described, they can beat the pre-existing teams at their game, at least for the SEO-outbound channels. Invest in those technologies and AI workflows.
- To keep up with this rapidly changing technology, you need a composable tech stack with an independent database to switch and replace parts of your workflow.
Is this same shift coming for AEs? Not completely. AEs are becoming puppets of AI models, recording and understanding an entire deal and recommending what to say and how to price the product.
While AI is better than humans at many things, people will still want to talk to people in some cases. Determine whether your customer or prospect prefers to buy from AI or a real person. That shift will happen gradually, with part of the market going to real humans and part to AI for the entire sales cycle.
The post How to Use AI-Powered Marketing to Get More Leads and Customers with Guillaume Cabane of HyperGrowth Partners appeared first on SaaStr.
via https://www.aiupnow.com
Amelia Ibarra, Khareem Sudlow
