How to reactivate old leads with AI

Start with a question your database cannot answer

An old enquiry tells you what someone wanted then. A useful reactivation call establishes whether that need still exists and what the next step should be.

Before writing an agent prompt, finish this sentence: “At the end of this campaign, our team needs to know ___.”

For postponed estimates, the answer might be whether the project is still planned. For an inactive candidate list, it might be whether someone wants a recruiter callback. Keep the first campaign focused enough that a response leads to a clear action.

Choose one segment with reliable context

Do not mix former customers, unanswered enquiries and postponed quotes into the same opening. They have different relationships with the business.

Segment	Useful context	First question	Useful outcome
Postponed estimate	Verified service and previous discussion	Is that project still something you are considering?	Requested quote review or updated timing
Old product enquiry	Verified product enquiry	Are you still exploring this?	Current need and requested callback
Inactive candidate	Verified role interest	Would you like a recruiter to contact you?	Availability and preferred callback time

Tip: if you cannot explain why a record belongs in the segment, leave it out of the first pilot. Review the applicable outreach permissions before including any contact.

Preserve a stable contact identifier

Use the original CRM record identifier in your working sheet. Keep it through preparation and results review so multiple attempts do not become multiple people in your report. Confirm which identifiers the product actually preserves before importing.

In the preparation sheet, record the enquiry source, verified context, last relevant interaction, timezone source, follow-up owner and outreach-permission reference. These are planning fields; use the current Kolz upload template for the actual import.

Tip: a phone number's area code alone may not establish someone's current timezone. Keep uncertain records separate until the scheduling information is resolved.

Write an opening that is easy to answer

Example to adapt after checking your context and outreach requirements:

“Hi, I'm the AI assistant for [business]. You previously asked us about [verified enquiry]. Is that still something you would like to discuss?”

Keep the context factual. Do not invent a recent interaction, a discount, an expiring deadline or a personal relationship. Start with one question; ask the next only after hearing the answer.

Prepare a truthful response for “Who are you?”, “Where did you get my number?” and “What does it cost?” If the agent lacks the answer, have it record the question for the team rather than make one up.

Test the branches, not just the happy path

Before using an external contact list, test these scenarios with your own team:

Scenario	What to check
Interested, asks for callback	The summary records the request and any stated time accurately.
Interested but refuses a callback	Interest is not converted into a claimed callback request.
“Maybe later”	The agent asks only necessary clarification and records uncertainty.
Wrong person	No additional enquiry details are disclosed; record is routed for correction.
Opt-out	The call ends appropriately and the record is excluded from future outreach. Verify the actual suppression process.
Pricing question outside instructions	No unverified quote or promise is produced.
Silence, interruption or voicemail	The result is not recorded as a qualified conversation.
Contradictory answer	The summary preserves the latest clear answer or marks it for review.

The accompanying test worksheet gives each scenario a place for observed behavior and a reviewer decision. A failed scenario should be fixed and retested before increasing campaign volume.

Define outcomes before looking at scores

Create a short review rubric that distinguishes these outcomes:

• Requested callback: the person explicitly asked or agreed to a human follow-up.
• Current interest: a present need was expressed, but no callback was agreed.
• Later: the person identified a future time or postponed the decision.
• Not interested: a clear negative response.
• Opt-out: a request to stop contact.
• Unknown: insufficient information to classify.

Keep technical call status separate from the business outcome. “Completed” may describe a call lifecycle; it does not establish interest or a sale.

Tip: manually review some high, medium and low scores. Looking only at the highest scores can hide false negatives. Include calls with missing transcripts in the review queue instead of treating missing evidence as success.

Match campaign size to follow-up capacity

A campaign can create more requested callbacks than a team can handle. Set a pilot size based on your ability to review conversations and complete follow-up, not just how many calls the system can place.

Record an owner and an agreed next action for every callback. Track both the requested callback and whether it actually happened. They are separate events.

Tip: use a small operational pilot to find mapping and conversation failures. Do not describe a small pilot as a statistically reliable conversion benchmark.

Report the funnel with explicit denominators

Metric	Definition
Contact reach rate	Unique contacts who answered ÷ unique contacts attempted
Callback request rate	Unique contacts requesting a callback ÷ unique contacts attempted
Callback completion rate	Completed callbacks ÷ callback requests
Customer conversion	Verified new customers ÷ unique contacts attempted
Cost per requested callback	Total defined campaign cost ÷ callback requests

If the denominator is zero, report the metric as unavailable. Always show the reporting window and whether follow-up outcomes have had time to mature.

Maintain usage charges, preparation time, review time and follow-up time separately. Use the actual billing rules and records when calculating costs. Reconcile modeled calculator outputs against a real campaign before presenting them as measured evidence.

Troubleshoot before expanding

Symptom	First checks	Useful response
Few answered calls	Number validity, scheduling information, attempt statuses and caller reputation	Inspect the cause before adding retries.
Answers but little current interest	Segment context and whether the question matches the original enquiry	Separate distinct enquiry types and revise one variable.
High scores with weak conversations	Rubric and transcript agreement	Tighten the outcome definition and repeat manual review.
Callback requests but few completed callbacks	Ownership, timing and whether handoff details were captured	Repair follow-up workflow before increasing volume.
Results look different by segment	Segment size, age, source and observation window	Report segments separately; avoid blending away the differences.
Large claimed revenue with little evidence	Duplicate attribution, existing opportunities and unverified sales	Require a recorded customer outcome and a defined attribution rule.

Test one change at a time

If you compare two openings, use comparable groups and keep scheduling, list source and follow-up process consistent where possible. Record the variant before launch. Decide what outcome you will compare before reading the results.

If the objective is to estimate incremental benefit, consider a comparable group that does not receive the reactivation treatment during the observation window. Track normal sales activity in both groups. Without a credible comparison, describe campaign-attributed outcomes rather than claiming the campaign caused every sale.

Bring the results back to the CRM

Review the output, match records using the stable identifier, and update the next action in your system of record. Verify the actual export/import workflow. A CSV workflow should not be described as a native automatic integration.

Keep outreach suppression separate from lead prioritization so a later import does not accidentally return an opted-out contact to the calling list.

For US campaigns, verify current rules covering the actual type of call, consent, opt-outs, scheduling and recording before launch. The operational steps here do not establish permission to call a particular list.

Start with a campaign you can explain

Prepare one clear segment, test the conversation branches, review the outcomes and assign follow-up. Expand when the workflow works and the evidence supports doing so.

Explore the Kolz database reactivation workflow and the existing cost and ROI calculator. Start with 100 free minutes.