Has AI made SDR personalisation too perfect to trust?

A few years ago, I ran an SDR workshop with my team in Dallas. We had an email problem: lots of emails, but little to no engagement or replies.

Like most SDR workshops, we talked about target accounts, buyer personas, value propositions, multi-touch campaigns, qualification and pipeline. All the usual areas that need to come together if you want to build a successful Sales Development team.

Then I decided to make the conversation a little more real. I opened my inbox to demonstrate what every person in that room was up against. I had labelled one folder “Prospecting”, with more than 1,500 unread emails that I had automatically routed once I recognised the source. But within that folder was a small handful that I had read, replied to or engaged with. I asked the team to look through them and answer one question: why did I read some, ignore most and engage with only a few?

That exercise mattered because it was their reality as SDRs. Every prospective customer they wanted to engage was facing the same thing: too much email, too little time and an inbox full of people claiming to have something important to say. The answer became clear very quickly.

Most of the unread emails were generic. Some had my name, used my company name, mentioned my job title or referenced a small piece of information found elsewhere. But they were still generic. They had taken a signal, dropped it into a template and hoped that would be enough to create relevance. The small handful I had read were different, timely and relevant.

They connected multiple dots: my role, the business, what was happening around us, a likely commercial challenge and a reason I might care at that point in time. It looked as though the SDR had done the work and earned my trust to engage. They had researched, thought and formed a point of view; they were in control of what they sent.

At that time, that level of personalisation was difficult to scale. It was messy, inconsistent, took time and relied on a resourceful human who could find the right information and connect it in a way that mattered. Fast-forward to today and the landscape has changed completely.

The evolution of personalisation

Enterprise SDR outreach has moved through four distinct stages.

1. Fully human research

This was the original model. An SDR would research the account, identify likely stakeholders, read annual reports, look at company announcements, study LinkedIn profiles, understand the market and try to form a relevant hypothesis. It was 100% human. It was often chaotic.

Some SDRs were fantastic at it. Others were not. Some messages had rough edges, inconsistent grammar and imperfect structure. But the best carried something that is becoming increasingly scarce: genuine human judgement. These were the messages copied a million times: an old-school version of AI at scale.

The SDR had taken the time to decide that this specific person, at this specific company, should receive this specific message.

2. Light personalisation

Then personalisation became more scalable: “Hi [First Name], I noticed you are the [Job Title] at [Company].” An SDR could make a thousand emails look personalised by inserting a few data fields. It was better than a completely generic blast, but buyers caught on very quickly. Adding somebody’s first name to a template is not personalisation. It is mail merge, and you would be surprised how much it is still used.

3. Single-signal trigger personalisation

The next evolution was more sophisticated. A company raises funding, a buyer changes jobs, a prospect publishes a LinkedIn post, or a business announces a new factory, product launch, expansion, acquisition, hiring drive or transformation program.

The SDR spots the signal, or increasingly the platform spots it for them, and a sequence is triggered. This was a significant step forward because there was a genuine reason to reach out. The problem is that when every SDR watches the same public signal, pushes it into an enrichment platform and sends the same kind of message, the signal itself becomes commoditised.

One person posts about hiring a new role; by lunchtime they have 20 emails from vendors saying they “loved the announcement” or asking whether they have thought about something. That is not relevance. It is a pack of wolves arriving at the same door at the same time looking for a meal.

4. Fully autonomous multi-signal personalisation

Now we are in a completely different world. AI can combine company news, hiring trends, technology data, market conditions, funding, job changes, intent signals, CRM activity, content engagement, buyer personas and account research into a detailed email within seconds. Five years ago, a message like that would have been considered a masterpiece. An SDR who could connect that many dots would have been held up as the benchmark for the whole team, promoted into leadership and perhaps written a book or two about it.

Today, it can be done automatically. That is where I think we need to stop for a moment, because the better AI gets at making outreach look personal, the more buyers are beginning to recognise the pattern: just like me several years ago in Dallas.

The problem with ‘too perfect’

I receive emails now that are highly relevant on paper. They mention a recent business change, connect it to my role, reference market trends and suggest a potential challenge. They are structured perfectly and beautifully written, on a level an English literature PhD student would be proud of. But I know almost immediately that nobody wrote them. They are too polished, complete and perfectly constructed: clearly too good to be true.

It is ironic that personalisation was supposed to help us stand out, but when every company uses AI to create multi-signal, hyper-relevant, beautifully structured outreach, that style of message becomes a new form of generic. Does it feel as though you have done your homework, or that an algorithm has done it for you?

It is a very different test, because if you do not truly understand my problem, why should I trust you to resolve it?

The business cost of autonomous perfection

This is not just a philosophical debate about whether AI sounds human. It has a direct impact on pipeline and may result in less meaningful engagement. Today, an email can be relevant and still fail to earn attention.

If a CRO, CMO, VP Sales or VP Marketing senses that a message was generated without human judgement, they may not reply, not because the problem is unimportant, but because the outreach did not earn enough trust to start a conversation in the first place.

For an SDR team, that becomes a pipeline problem. Sending more relevant emails does not automatically create more quality conversations, and more data does not necessarily generate better buyer interest.

More activity, less learning

When AI runs the research, decides the target, writes the outreach, sends the sequence and triages the reply, what is the human SDR learning? This is your talent bench of future sales stars. Sales Development needs to be more than an automated messaging factory. It should be the front end of your revenue engine and a training ground for future AEs.

• Understand a target account.
• Form a commercial hypothesis.
• Identify and engage multiple personas.
• Ask questions that uncover Pain.
• Establish real business Impact.
• Understand Critical Events.
• Start mapping the Decision process.

That is SPICED, from Winning by Design, in practice. When we allow AI to remove the human from every one of those judgement moments, we might create more automated activity, but we do not necessarily create better enterprise sellers. You may not feel it today, but when you promote a superstar SDR into an AE role and discover they do not have the judgement you need, that is on you.

Reduced trust in the SDR function

The worst outcome is not a lower email open rate. It is that SDRs become known internally and externally as people who operate automated outreach machines. That will not build future Sales Executives.

That is the culture you want people to describe. When an SDR has no ownership of the message, no ability to apply judgement and no responsibility for the commercial hypothesis, the work starts to feel less meaningful. If the buyer feels the same thing, trust disappears on both sides, and you end up in the folder of 1,500 unread emails.

AI should support SPICED, not replace it

AI has a major role to play in Sales Development. This does not mean returning to an hour of manual research for every account, writing every email from scratch or refusing to use the tools available to us. That is unrealistic. But we do need to be clear about where AI helps and where the human must remain accountable.

Situation: AI finds company, market, technology, hiring and persona signals. The SDR decides which facts matter and creates a relevant account hypothesis.

Pain: AI suggests likely challenges based on industry and account context. The SDR confirms the buyer’s real pain in their own words.

Impact: AI highlights possible revenue, cost or experience implications. The SDR establishes what really matters to that buyer, both rationally and emotionally.

Critical Event: AI spots public triggers, deadlines and business events. The SDR tests urgency: what happens if you miss that date?

Decision: AI maps likely stakeholders and public organisational structures. The SDR learns how the organisation makes decisions and builds the real buying group.

AI can create a starting point, give an SDR a richer view of the account, help identify who to contact, show why now might matter and suggest useful questions. It can even highlight whether the account deserves outreach. It cannot honestly claim a buyer has a Pain before the buyer has said so, establish Impact on behalf of the customer, or build trust simply because an email contains more signals than a human could find in ten minutes.

The answer is not fake imperfection

There is one response to this problem that I do not agree with: deliberately making emails messy, inserting bad grammar, random mistakes or manufactured ‘human’ imperfections to fool the buyer. That is just another form of automation, not to mention an unprofessional breach of the one thing you need to establish: trust. Buyers will spot that too.

The goal is not to make AI-generated outreach look worse. It is to make the SDR accountable for the parts that require judgement. Call it AI-assisted, human-accountable outreach.

AI does the heavy lifting:

• Account research.
• Data enrichment.
• Signal detection.
• Initial draft generation.
• Sequencing support.
• CRM updates.
• Reply categorisation.
• Workflow reminders.

The SDR owns the critical moments:

• Selecting the accounts that genuinely deserve attention.
• Checking the facts.
• Deciding whether the signal is relevant or merely interesting.
• Adding a real observation or point of view.
• Ensuring the tone is appropriate.
• Taking ownership when a buyer replies.
• Converting a promising conversation into SPICED-quality discovery.

That is the real human-in-the-loop model, not a human clicking ‘approve’ on 500 messages they have not read, written or even understood, but a human involved at the decision points where credibility is won or lost.

A better enterprise SDR play

For enterprise accounts and senior personas, I would build a simple play around this. Let AI create the hypothesis by surfacing:

• Why this account fits the ICP.
• Which business event or market condition may be relevant.
• The likely personas involved.
• A potential problem worth exploring.
• The data points that need validating.

Then make the SDR test the hypothesis. Before anything goes out, the SDR should be able to answer:

• Why this account?
• Why this person?
• Why now?
• What do I genuinely believe may be relevant to them?
• Would I be comfortable saying this exact message in a face-to-face conversation?

If they cannot answer those questions, the message should not be sent. Keep the first message focused on one real observation, one relevant connection and one low-friction question, not six data points, a full account plan or a thesis written by an AI model trying too hard to prove it researched the company.

The objective is not to demonstrate how much you know. Two minutes of AI research will probably tell you more about a company than the person knows you are about to reach out to today. The objective is to give the buyer a credible reason to respond, and that starts with trust.

AI needs to be celebrated through a measure of quality, not just automation. If you are celebrating AI because it sends more messages, you have gone back to the future: send more, think less.

Measure whether it improves:

• Positive response rate.
• First meetings held.
• Sales-accepted opportunities.
• SPICED completeness at handover.
• Multi-persona engagement in enterprise accounts.
• Opportunity progression.
• Pipeline created and converted.
• Buyer and AE feedback on opportunity quality.

When AI increases send volume but reduces trust, response quality and downstream conversion, it is not helping your pipeline engine or tomorrow’s talent bench. It is simply making the wrong activity cheaper and more commoditised.

Where are you taking this?

The future of SDR outreach will not be fully human or fully autonomous. It is a better division of labour. AI will continue to improve at research, pattern recognition, data connection and draft generation, and that is a good thing.

The scarce asset in enterprise Sales Development will become human judgement. It always has been, but the headlong rush into automation has lost sight of that.

An SDR who can say that about an AI-generated message will stand out, because they are building the critical skills that made SDR one of the most important roles in the company. If they can turn a machine-generated hypothesis into a real buyer conversation, they will stand out even more.

So my question is not whether we should use AI. It is this: at what point in your SDR outreach process does a real human still have to own the message, and are you honest with yourself about where that line is?