The six steps to empowering a sales team with AI: lead enrichment, scoring, outreach, meeting preparation, follow-up and measurement, in field-tested order.
Link copied
The sales director of a B2B SaaS company called me two months ago: "Yüce, we send 5,000 outbound emails a month with AI and conversion has dropped to 0.3 per cent. It was 1.7 per cent when we did it manually. AI is killing us."
I read the sentence backwards. What was killing the team was the poor use of AI far more than AI itself.
With AI, sales either makes a big leap or slides into brand erosion. The line between the two is discipline. Teams that skip the structure pull back within a few months, saying "AI didn't suit us". The fit problem sits in the setup far more than in the AI.
Salesforce's 2026 State of Sales report says something clear: sales teams that integrate AI properly see a 67 per cent productivity gain. Teams that integrate it badly end up net negative. The single variable between them is methodology. The 6 steps below are the backbone of that methodology.
In old-style sales, an SDR spends 2 hours on LinkedIn, builds a lead profile and types it into the CRM. That yields 8 to 10 leads a day.
With an AI layer such as Apollo, Clay or similar, that time drops to 8 minutes. You start with a company name and a LinkedIn URL and AI does the enrichment. The CEO's name, the funding round, a summary of the last 6 months of news, the technology stack and headcount changes are gathered into a single table.
The pattern I see in B2B outreach teams is this: the full profile of 1,000 to 1,500 target companies comes together in three to five days. Manually, the same job used to take 6 weeks.
More critical still is profile depth. A table richer than any human could compile takes shape. An SDR naturally checks 12 data points; AI collects 47 in parallel.
The classic enrichment mistake: data AI has collected gets embedded into emails without verification. The data can be stale, or the wrong person can be tagged. Human sign-off must remain a step in the process.
The second step is prioritising the leads waiting in the queue. Classic scoring rules fall short because they are static. Company size, sector and revenue band do not point the way on their own. AI scoring watches behaviour and keeps updating itself.
A good AI scoring layer synthesises more than one signal. Website visit depth, time spent on the pricing page, engagement with marketing content, the warmth of past SDR correspondence and signals in the company's last 90 days of news all feed into one model. A new CTO appointment, fresh funding or entry into a new geography shifts the score instantly.
According to data Workday published in early 2026, sales teams working with AI scoring lift close rates by 37 to 40 per cent. The reason is plain: the SDR talks to the right lead at the right time.
The most common mistake in B2B sales teams is skipping this step. The CRM's scoring was set up 5 years ago and has never been updated. An AI scoring layer goes on top of that CRM within three to four weeks and shows results within 60 days.
This is the most abused step. The moment a sales director says "we send 5,000 emails with AI", something bad is already happening.
Correct AI outreach works with fewer, sharper touches. The SDR approaches 15 high-scoring leads a day, working from the personalised message draft AI has prepared. AI personalises, the SDR checks, then sends.
A message needs 4 elements. First, a specific trigger: an event at the lead's company within the last 30 days. Second, relevance: the connection between that event and your offer. Third, a one-sentence value proposition. Fourth, a low-friction question. A closing in the tone of "what is your current thinking on this" rather than "let's talk for 15 minutes".
Platforms such as Apollo, Outreach.io and Salesloft offer AI outreach layers. Using them demands a quality threshold. A message being human-like is no longer enough; it has to be specific. If you are keeping messages that open with "Dear Sir or Madam", your team is producing email volume rather than sales.
A large share of the productivity gain in the Salesforce report comes from this step. In the old routine, an SDR wanders around for 30 minutes before a meeting, opens the last interaction and browses the website.
AI-driven preparation takes 5 minutes. An hour before the meeting, AI sends a preparation pack. A summary of the customer's interactions over the last 6 months, sector benchmark data, likely objections, notes from the previous meeting and the contact's latest LinkedIn activity are collected on a single page.
The pattern I have seen in large group sales functions: 40 to 50 sales representatives average 6 meetings a week.
Saving 25 minutes of preparation per meeting frees up thousands of hours of capacity a year. When that time is redirected into sales conversations, a double-digit percentage jump in quarterly revenue becomes possible.
This step is the hidden backbone of sales. Most teams talk about lead generation and skip past follow-up. Yet half of all unclosed deals are lost to poor follow-up.
An AI follow-up layer suggests, for every open lead, when to make contact, with which message and through which channel. When the SDR opens the CRM in the morning, today's 12 follow-up actions sit in a list. The suggested message for each one is ready. The SDR approves and sends.
A more advanced layer offers a pipeline health score. It tracks the time each deal spends at each stage and alerts the SDR when it spots an anomaly: this lead has been waiting at stage two for 21 days, and against historical data that is a deviation.
The final step is the one most teams skip. You set AI up, you run it, you never read the numbers. Or you read the wrong numbers.
The right measurement metrics are well defined. Weekly qualified meetings per SDR, open and reply rates of AI-assisted versus non-AI messages, close rates of meetings opened through scoring, usage rate of the preparation pack and the team's AI satisfaction score are all tracked.
Do not accept the 67 per cent productivity headline at face value. The number will come out differently in your team. What matters is the direction of the trend. You measure a baseline for the first 90 days, then take the post-AI measurement over the next 90. The delta is yours.
Step | Speed of Result | First 90-Day Investment |
Lead Enrichment | High | Apollo/Clay subscription |
Lead Scoring | Medium-high | CRM AI layer |
Outreach Automation | Medium (high risk) | Outreach.io / Salesloft |
Meeting Preparation | Very high | In-house AI agent |
Follow-up and Pipeline | High | CRM add-on |
Measurement and Feedback | Continuous | Dashboard setup |
Start with the first and fourth steps. The lowest risk and the fastest gain sit there. Scoring comes second, then follow-up. Outreach automation stays last, because poor calibration at that step creates the highest brand cost.
Working with a large number of sales teams in the field, I noticed something: 92 per cent of those who said "it didn't suit us" had done one of three things. They had started with outreach automation; wrong order. They had imposed AI on the SDRs from the top down without asking; adoption resistance. Or they were measuring nothing, and decisions were made on feel.
None of them had the infrastructure to claim "AI didn't work". I told them they had decided without measuring. They went quiet.
The 6 steps to getting results from AI in sales are lead enrichment, lead scoring, outreach automation, meeting preparation, follow-up and measurement. Salesforce reports a 67 per cent productivity gain; that figure holds for teams that apply the steps in the right order.
Start with enrichment and meeting preparation. Leave outreach automation until last, because it carries the highest brand risk.
The upstream link that determines demand quality is marketing. The five ways to get results from AI in marketing explains the same discipline on the marketing line. For a picture of what AI agents look like at operational scale, read the Klarna case study analysis.
If you would like these 6 steps delivered to your sales team as a live case story on stage, browse Speaker Agency's artificial intelligence speakers and get in touch. Our guide on how to choose an AI keynote speaker will help you shortlist the right name.
It acts as an assistant rather than a replacement. SDR and AE roles stay; a single SDR's output triples. According to Salesforce, 60 per cent of sales teams will be working in hybrid mode with AI agents by 2027. Teams grow their capacity instead of shrinking.
Apollo, Clay, Outreach.io and Salesloft sit in the 80 to 200 dollar range per user per month; the total budget for a 10-person sales team takes shape accordingly. ROI starts within the first 90 days; Apollo's average customer reports a multiplying return within a few months.
Language models have made notable progress over the last 18 months. Claude and GPT-4.5 write outreach at human quality. Cultural nuance still calls for human review. Respectful language, the correct salutation and sector jargon belong on the checklist. AI writes, the SDR approves and sends.
Most modern CRMs (HubSpot, Salesforce, Pipedrive) offer AI scoring integration. On a legacy system a middleware layer is needed; setup takes 4 to 6 weeks. Most sales teams can build an AI layer on top of the existing system without moving to a new CRM.
Used wrongly, absolutely. A team firing off 5,000 generic emails pushes its own domain into spam. Used correctly, an SDR sends 15 to 25 high-scoring, personalised messages a day and every message passes human review. Automation works for scale, personalisation works for quality; the two travel together.
After Dinner Speakers
Contents
