If you run sales or customer success for more than five minutes, you learn this annoying truth.
Most of your day is guessing.
Who should I call first. Who is most likely to answer. Who is just “checking in” forever. Who is actually about to buy but hasn’t said it yet.
And sure, you can go with gut feel. Or “hot leads” from forms. Or whatever your CRM says is “high priority” (which, half the time, is just the last person who opened an email).
But AI is really good at one thing humans struggle with.
It can look at messy past behavior, across hundreds or thousands of accounts, and spot the patterns that usually come right before a purchase.
Not vibes. Not hope. Patterns.
This is how AI looks at past patterns to tell you who to call today. In plain language. And in a way that actually leads to more closed deals, not a fancy dashboard you never use.
The basic idea: people leave buying “footprints”
Customers rarely wake up and buy out of nowhere.
They do stuff first.
They visit certain pages. They ask certain questions. They open emails at certain times. They log in more often. Or they stop logging in. They add teammates. They download an integration guide. They request pricing twice. They forward your proposal. They reply faster. They suddenly care about onboarding.
Those are footprints.
A human can notice a few of them. But once you have more than, say, 50 active deals. It gets fuzzy. You miss things. Or you chase the loudest lead, not the best one.
AI doesn’t “understand” customers like a person does. But it can count and compare behavior at scale. It looks at the footprint trail and says:
“This trail looks like the trails that usually end in a purchase.”
And then it ranks your accounts or leads accordingly.
What AI is actually doing (without the math lecture)
At a high level, AI prediction for “who will buy next” works like this:
1. You show it your history
Past leads and customers. The ones who bought, the ones who didn’t, the ones who churned, the ones who expanded.
2. You show it what they did before the outcome
Activities and attributes leading up to the result. Emails, calls, meetings, site visits, product usage, support tickets, industry, company size, deal size, time in pipeline, all that.
3. It finds combinations that repeat
Not just one signal like “opened email.” The AI looks for patterns like: opened 3 emails in 5 days, visited pricing page twice, added 2 users, asked a security question, had a meeting scheduled within 7 days, and deal age under 21 days. When those show up together, the close rate jumps. The AI learns that.
4. It scores current leads based on similarity
Today’s accounts get compared to those past patterns. The AI spits out a score or a rank. Basically: “This one looks like buyers.” “This one looks like maybe later.” “This one looks like a time sink.”
That’s it. That’s the whole thing.
It’s not magic. It’s pattern matching at scale.
The signals AI tends to care about (aka what it learns from)
Different businesses have different signals, but most prediction systems end up pulling from a few buckets.
1) Fit signals: “Are they the right type of customer?”
These don’t change daily, but they matter.
- industry
- company size
- job titles involved
- region
- tech stack
- whether they match your best customer profile (ICP)
A lead that behaves like a buyer but is a terrible fit can still waste time. So AI typically mixes fit with behavior.
2) Intent signals: “Are they showing interest right now?”
This is where the “who to call today” part gets real.
- pricing page views
- demo page visits
- comparison pages (“yourbrand vs competitor”)
- downloading a buyer guide
- replying to emails quickly
- booking time on the calendar
- repeat visits from the same company IP
- multiple stakeholders engaging
Intent signals are usually time sensitive. Yesterday matters more than 6 months ago.
3) Engagement signals: “Are they leaning in, or just browsing?”
AI watches how people engage, not just whether they showed up once.
- frequency of touchpoints
- time between touches (momentum)
- number of stakeholders involved
- meeting attendance rate
- length of sales cycle so far
- whether the deal is moving stages or stuck
A common pattern is momentum. Buyers build momentum. Tire kickers spike once, then vanish.
4) Product signals (for SaaS): “Are they getting value?”
If you have a free trial, freemium, or product led motion, usage is huge.
- number of active days in last 7
- key feature usage (the “aha” action)
- number of projects created
- invites sent
- integrations connected
- errors experienced (and whether they got resolved)
- support tickets raised
A weird but real pattern: sometimes support tickets increase right before purchase. Because real buyers are trying to make it work in their environment.
5) Commercial signals: “Do they behave like buyers in the pipeline?”
AI can also learn from the sales process itself.
- stage changes
- proposal sent
- procurement involved
- security review initiated
- legal redlines started
- discount requested
- contract sent
Some of these are obvious to humans. But AI can learn the timing. Like “if security review starts within 10 days of first demo, close rate is 2x.”
So how does that translate into “who should I call today”
Here’s the practical version.
Instead of opening your CRM and sorting by “Last contacted” (which is basically a to do list, not a revenue list), you open a prioritized view like:
- High likelihood to buy soon
- Medium likelihood
- Low likelihood
- At risk / going cold
- Likely upsell / expansion
And each account has a few “reason codes” attached. Not always, but the better systems do this. Things like:
- “3 pricing page visits in 48 hours”
- “New stakeholder engaged”
- “Usage increased 40% week over week”
- “Proposal opened 5 times”
- “Trial ends in 3 days and feature X used”
This is the key part. The rep doesn’t just see a score. They see why.
Then your call list becomes simple:
- Call the high likelihood leads with fresh intent signals first.
- Call the at risk deals that are showing early warning signs.
- Spend less time “checking in” on low likelihood leads unless there’s a reason.
It’s not about replacing reps. It’s about telling them where their effort has the highest return, today, not someday.
A quick example (what a pattern can look like)
Let’s say you sell a B2B tool with a demo led sales process.
AI might learn something like this from your last 2,000 deals:
Deals that closed within 30 days often had:
- 2 or more people from the company visiting the site within a week
- at least 1 pricing page visit
- demo booked within 5 days of first touch
- a follow up email reply within 24 hours
- a security or integration question asked before the second meeting
So today, when an account does those things, the AI pushes it up the list.
Meanwhile, accounts that do this:
- download one ebook
- no follow up engagement
- only one person involved
- long gaps between responses
They go down the list. Not because they’re bad people. Just because your history says they rarely buy soon.
What you need for this to work (and what people mess up)
AI needs decent inputs. Not perfect. But usable.
Here’s what usually matters most:
Clean outcomes
You need to know what “won” and “lost” actually mean in your CRM. If everything is marked “closed lost” with no reason, you can still predict, but it’s noisier.
Consistent activity tracking
If reps log calls randomly, or half your meetings happen outside the calendar system, AI won’t see the full picture.
Enough volume
If you have 20 deals a year, AI won’t have much to learn from. But once you have hundreds of past opportunities, patterns become clearer fast.
A feedback loop
This part gets ignored. Reps should be able to say “this was wrong” or “this lead was actually hot.” That feedback improves the model or at least your rules over time.
And also. If your team doesn’t trust the output, they won’t use it. So you need transparency. The “why this is ranked high” notes matter a lot.
What results should you expect (realistic ones)
AI lead scoring and next best action systems usually don’t turn a bad product into a good one. Or fix a broken sales process overnight.
But they do a few very specific things well:
- Higher connect to close efficiency: reps spend time on people who are actually in motion
- Faster follow up on real intent: you catch buyers while they’re warm
- Less pipeline delusion: fewer “we’ll see” deals get treated like real forecasts
- Better timing: you call when the customer is already leaning forward
In plain terms. More calls that matter. Fewer calls that feel like shouting into the void.
How to start without making it a giant project
If you want to use AI to decide who to call today, start small:
- Pick one motion New inbound leads. Trials. Expansion. Whatever. Don’t try to score everything at once.
- Decide the outcome “Booked a demo,” “became sales qualified,” “closed won,” “expanded.” Pick one.
- Feed it the last 12 to 24 months Enough to learn patterns, not so much that your go to market motion has changed completely.
- Make the output usable A ranked list in the CRM. With reasons. Not a separate tool nobody opens.
- Measure one metric Speed to first call on high intent leads. Or conversion rate of top ranked leads vs the rest. Keep it clean.
The whole point
AI is not here to “predict the future” like a fortune teller.
It’s doing something more boring. More useful.
It looks at your past customers, figures out what buying tends to look like in your world, then taps you on the shoulder and says:
“Call this one first. This is the pattern that usually turns into revenue.”
And that is how you stop guessing.
FAQs (Frequently Asked Questions)
Why is guessing a big part of sales and customer success, and how can AI help?
In sales and customer success, it’s common to spend much of the day guessing who to call first, who’s likely to answer, or who is ready to buy but hasn’t said so. AI helps by analyzing messy past behavior across many accounts to spot patterns that usually precede a purchase, allowing you to prioritize leads based on data rather than gut feelings.
What are buying “footprints” and how does AI use them?
Buying footprints are actions customers take before making a purchase, like visiting pricing pages, asking questions, opening emails at certain times, or requesting pricing multiple times. AI tracks these footprints across numerous deals to identify patterns that typically lead to purchases, helping sales teams focus on the most promising leads.
How does AI predict which leads are most likely to buy next?
AI prediction involves four steps: 1) learning from your historical data of past leads and customers; 2) analyzing what activities or attributes led up to buying outcomes; 3) finding recurring combinations of signals (like multiple email opens plus pricing page visits); and 4) scoring current leads based on similarity to these successful patterns, ranking them by likelihood to buy.
What types of signals does AI analyze to score leads?
AI analyzes several categories of signals: Fit signals (customer profile like industry and company size), Intent signals (current interest shown via page visits or demo bookings), Engagement signals (interaction frequency and momentum), Product signals (usage patterns in SaaS products), and Commercial signals (sales process behaviors like proposal sent or contract negotiation). Combining these helps accurately prioritize leads.
How do intent signals influence who sales should call today?
Intent signals reveal if a lead is showing active interest right now through actions like visiting pricing or demo pages, downloading buyer guides, replying quickly to emails, or booking meetings. Because these behaviors are time-sensitive—recent activity matters more—AI uses them to identify leads with the highest likelihood of purchasing soon for immediate outreach.
What practical benefits does AI-driven prioritization bring compared to traditional CRM sorting?
Instead of sorting leads by last contacted date—which often results in a simple task list—AI-driven prioritization ranks accounts by likelihood to buy soon, medium likelihood, low likelihood, at risk/going cold, or potential upsell opportunities. This approach focuses sales efforts on revenue-generating activities by highlighting the best prospects with clear reasons for their ranking.
