# AI SDRs are a waste of time. For now. URL: https://www.coltasa.com/insights/ai-sdrs Author: Colin Constantine Published: 2026-10-05 Category: Pipeline > Why AI SDRs fall short in enterprise sales today, what would need to be true for that to change, and where AI actually helps your team. Photo: [Alex Knight](https://www.pexels.com/photo/high-angle-photo-of-robot-2599244) on Pexels Every few weeks a founder asks me whether they should replace their SDR team with an AI SDR. The pitch is attractive: an agent that researches prospects, writes emails, handles replies and books meetings around the clock, at a fraction of the cost of a person. My answer, today, is no. Here is why, and what would need to change for that answer to change. ## What an SDR is actually for The job of an SDR is not to send emails. It is to start conversations with the right people at the right companies, and to qualify whether there is a real opportunity worth an account executive's time. Sending is the easy part. The hard parts are judgement and trust: knowing which account is worth pursuing, what to say to this particular buyer, and when to stop. AI SDRs automate the easy part extremely well. They mostly fail at the hard parts. ## Why they fall short today They add to the noise. When every vendor can deploy an agent that sends thousands of personalised emails, buyers learn to ignore all of them. Many would rather skip sellers entirely: in Gartner's 2026 survey, [67 percent of B2B buyers said they prefer a rep-free buying experience](https://www.gartner.com/en/newsroom/press-releases/2026-03-09-gartner-sales-survey-finds-67-percent-of-b2b-buyers-prefer-a-rep-free-experience). The tactic stops working for everyone, including the companies using it. Volume is not a moat when everyone has it. The personalisation is shallow. Mentioning a prospect's recent post or funding round is not insight. Senior buyers can tell in one sentence when an email was assembled from a data feed. It signals the opposite of what you intend: that nobody at your company thought about them. They damage your brand quietly. A human SDR who sends a bad email embarrasses themselves. An agent that sends ten thousand mediocre emails trains your entire market to associate your name with spam. That cost does not show up on the dashboard, but it shows up later when your AEs try to get a meeting. They cannot qualify well. Real qualification is a conversation. It involves reading hesitation, asking a follow-up that was not in the script, and recognising when "send me more information" means "go away". Agents book meetings, but many of those meetings are with people who were never going to buy. Your AEs pay for that in wasted calls. They hurt deliverability. High volume, low engagement and lots of new sending domains is exactly the pattern spam filters are designed to catch. Google's [sender guidelines](https://support.google.com/mail/answer/14229414?hl=en-GB) ask senders to keep reported spam below 0.1 percent and never reach 0.3 percent. Teams end up buying domains and warming inboxes to keep the machine running, which is effort spent fighting a problem they created. The economics look better than they are. Compare cost per qualified opportunity, not cost per email or per meeting booked. Once you include unqualified meetings, burned accounts and the time spent managing the tool, the gap with a good human SDR is much smaller, and often goes the other way. ## What would need to be true I do not think this stays true forever. I would change my view if AI SDRs could reliably do the following: - Produce genuine insight about a prospect's business, the kind a buyer would forward to a colleague, not a rephrased LinkedIn summary. - Hold a qualifying conversation that a senior buyer would rate as useful, including knowing when to stop. - Improve conversion to [qualified pipeline](https://www.coltasa.com/insights/pipeline-coverage), measured over a full sales cycle, not just meetings booked. - Operate at low volume with high accuracy, rather than relying on sending more to hit targets. - Do all of this without buyers feeling deceived when they learn they were talking to software. Some of this will arrive. When it does, I will happily recommend these tools. Until then, the burden of proof sits with the vendor. ## What to do instead Use AI where it makes your people better, not where it replaces the part buyers value. - Research: let tools gather company news, org charts and trigger events so your SDRs start with context. - Preparation: use AI to draft account briefs and call plans that a human then edits. - Admin: automate CRM updates, call notes and follow-up reminders. - Coaching: review recorded calls to spot where conversations stall. Then point your human SDRs at fewer, better accounts and hold them to a quality bar: meetings that turn into real opportunities. There is a practical weekly rhythm for this in [the contract between buyers and sellers is broken](https://www.coltasa.com/insights/signal-over-noise). ## If you are going to try one anyway Run it as a controlled test. Give it a separate segment, a separate domain and a fixed time window. Measure qualified opportunities and pipeline value against a human-led control group over at least one full sales cycle. Read a sample of what it sends every week. If you would be embarrassed for your best customer to receive it, stop. Buying decisions in enterprise sales are still made by people who want to deal with people who understand their problem. Build your pipeline around that, and use the tools to help. ## Sources and further reading - [Gartner (2026): Sales survey finds 67% of B2B buyers prefer a rep-free experience](https://www.gartner.com/en/newsroom/press-releases/2026-03-09-gartner-sales-survey-finds-67-percent-of-b2b-buyers-prefer-a-rep-free-experience) - [Google: Gmail email sender guidelines FAQ](https://support.google.com/mail/answer/14229414?hl=en-GB)