DOES AI OUTBOUND ACTUALLY WORK? HERE IS WHAT OUR PIPELINE DATA SAYS
By Ben Fildes · 10 July 2026 · 4 min read
There is a question doing the rounds in the investment community this week. Sequoia partner David Cahn calculated that the AI industry will need to generate $3 trillion in revenue just to justify its 2026 infrastructure spend. Hyperscalers are pouring $1.5 trillion into chips and data centres. To break even on that, the technology has to produce extraordinary commercial output. Fast.
You might think this is a Wall Street problem, not yours.
It is yours. Because your prospects have read the same headlines. They are not asking whether OpenAI will hit a revenue target. They are asking whether your AI pitch deserves more than thirty seconds of their attention. "Does this actually work?" is now the first objection in every discovery call.
Here is the honest answer, from someone running it live.
Why the ROI question matters for founder-led businesses
The global AI ROI debate has filtered down into the conversations founders are having with their own buyers. If you sell B2B services and you are using AI anywhere in your business, you will face this question before the end of this quarter.
The trap is answering it the wrong way. Citing industry averages ("87% of sales teams now use AI") does nothing. Gesturing at product demos does nothing. The only thing that moves a sceptical founder is a specific number tied to a specific outcome they recognise.
The good news: if you are running AI-assisted outbound on LinkedIn, those numbers exist in your own system. The bad news: most people are not measuring them.
What we see in our own data
We run a LinkedIn outreach system called SalesOS. It runs on the founder's own LinkedIn account. Every message is reviewed and approved before it sends. Nothing goes out on autopilot.
Across our own LinkedIn account, the system has opened around 700 conversations since June and booked 20 meetings, 8 of them in the last 30 days. Zero cold email, zero paid advertising, no outsourced SDR team, and every message human-approved before it sent.
The conversion path looks like this: of every 100 engaged contacts, roughly 9 reply and signal genuine interest. Of those interested, around 30 book a call. The 9%/30% pair is the number we now use to plan capacity for new clients before we onboard them.
In the last 30 days, the AI made 296 separate funnel stage assessments across active conversations. 108 of those contacts moved forward in the pipeline as a result of those decisions, with the right follow-up message going to the right person at the right moment. Every single one was reviewed before sending.
Two numbers worth sitting with: 296 decisions. 108 advancements. All inside one founder's working week, with none of the spreadsheet admin those decisions would normally require.
What to do about it: four concrete moves
1. Define your unit of success before you start. Meetings booked per hundred conversations opened is the cleanest metric for LinkedIn outbound. Not open rate. Not click rate. Meetings. If you cannot trace back from a booked call to the outreach that opened the conversation, you cannot measure anything.
2. Keep the human in the loop, and log it. Every approval or edit you make to a staged message is a data point. Over time, the pattern of your edits tells you which message angles are landing in reality versus theory. Our pipeline data shows that before/after/bridge framing dominates the drafts that make it through without edits. That is not a creative preference. It is a measurement outcome.
3. Separate volume from quality early. A high connection acceptance rate and a low reply rate mean you are reaching the right companies but with the wrong message. A low acceptance rate and a high reply rate from those who do connect means your targeting is too narrow but your copy is strong. These are different problems with different fixes.
4. Set a 30-day baseline before you change anything. The temptation once the system is running is to adjust angles, targeting, and message cadence all at once. Resist it. The only way to know what is working is to run one configuration long enough to get a statistically honest read. Thirty days is the minimum. Twelve is enough to know whether you have a pipeline, not whether you have a system.
The bigger picture
The $3 trillion question is real. The infrastructure spend on AI is extraordinary, and not all of it will generate returns. But for founder-led B2B businesses, the ROI calculation is much more immediate.
Did the meeting happen? Did the conversation start? Did the right person reply?
If the answer is yes, consistently, across multiple accounts and multiple months, then AI is paying for itself with room to spare. The infrastructure debate is a macro problem. Your pipeline is a micro one. Solve the micro one first.
How SalesOS fits
SalesOS is the system we built to produce exactly these numbers for founder-led businesses in services, recruitment, and B2B SaaS. It runs on your LinkedIn account, generates pipeline in your name, and keeps you in the approval seat throughout.
If you want to see what the numbers look like for your specific market before committing, book a call here.
One new client in 90 days or every penny back.
Founder of Neon Gorilla. First Class BA in Marketing and an MSc in Enterprise and Innovation (Distinction) from Keele. Previously co-founded Beast Biltong with Eddie Hall, stocked in 2,000+ stores. Everything here is written from our own campaign logs, not theory.
More about Ben →AI Sales is one of three engines we run on your own accounts, you approve every move. We owe you 1 client in 90 days, or it's free until you get one.