The qualification call went well. The AI SDR identified a 140-headcount logistics company with a remortgaging need, confirmed budget, verified the decision-maker was on the line, and triggered a handoff. The rep received a Slack message: "Qualified lead — call [Name] at [Number]."
The rep called back 47 minutes later. The prospect had already booked with a competitor who'd called within 10 minutes of the same LinkedIn ad click.
The agent did its job. The handoff did not.
This is the silent revenue leak in most UK AI outbound programmes. The qualification rate gets measured. The booked-call rate gets measured. The gap in between — the context gap between when the AI stops and the human starts — is where deals disappear without anyone noticing or being accountable.
What actually breaks at the handoff point
Four failure modes, in order of frequency:
1. Latency. The handoff triggers, but the AE doesn't act within the prospect's patience window. For an inbound enquiry or a just-qualified warm call, that window is 15–30 minutes. After 30 minutes without contact, conversion rate drops by an estimated 40–60% based on InsideSales (now Xant) research on speed-to-lead — a finding consistently reproduced in UK B2B contexts in our own campaign data.
2. Context loss. The AE knows the prospect's name and number. They don't know what the prospect said they wanted, what objection they raised, or what the agent promised. The AE opens with "I understand you spoke with someone from our team earlier" — which sounds like nobody briefed them, because nobody did.
3. Duplicate outreach. The handoff fires, the AE calls back, the prospect says "I already spoke to someone who said they'd send a proposal." Meanwhile the sequence is still running and an email fires tomorrow. Two channels, neither coordinated.
4. No SLA, no escalation. The handoff Slack message arrives while the AE is on another call. It sits unread for 3 hours. There's no system to escalate, no fallback, no metric that catches it.
The numbers aren't unique to our clients. Gong.io's analysis of 100,000+ B2B sales calls found that deals where AEs had prior call context in their CRM before the first outreach closed at 22% higher rates than deals where the context was absent. The context isn't nice to have — it's a measurable revenue variable.
A counterpoint worth considering: Salesloft's 2024 State of Sales Development report argues that over-engineering the handoff brief can actually slow AE response time — reps who receive extensive pre-call documentation take longer to review it and respond later, not earlier. The evidence suggests a 50–100 word summary outperforms both "here's the transcript" (overwhelming) and "here's the name and number" (useless). That aligns with what we see in practice.
The four data fields every AE needs before saying hello
Strip back the context requirement to what an AE actually needs in the first 30 seconds of a call-back to convert:
{
"contact_name": "Sarah Okonkwo",
"company": "Apex Logistics UK",
"qualification_status": "qualified",
"intent_signal": "Requested pricing for 10-agent outbound calling setup",
"call_summary": "Sarah confirmed budget (£15–20k), currently using manual dialling across 6 reps, wants to reduce cost and increase coverage. No existing voice AI. Decision timeline: 30 days.",
"disqualifying_flags": [],
"handoff_timestamp": "2026-06-19T14:23:00Z"
}
Everything else — tech stack detail, full call transcript, objection analysis — goes in the CRM record for the AE to review before a follow-up call, not before the initial call-back. Information overload at handoff time slows the AE down rather than helping them.
The intent_signal field is the most important. It's the specific thing the prospect said or did that triggered the handoff. An AE who knows "she asked for pricing" opens very differently than one who just knows "qualified lead". The first 10 words of the call-back either confirm or contradict what the prospect thinks is happening. Get this field right.
Handoff trigger design: when does the SDR stop and the AE start
The handoff trigger is a condition, not just an event. Defining it precisely determines how many qualified leads get passed versus how many false-positives the AE has to sift through.
A working trigger schema:
handoff_trigger:
conditions:
- field: qualification_score
operator: gte
value: 75
- field: budget_confirmed
operator: eq
value: true
- field: decision_maker_confirmed
operator: eq
value: true
require_all: true
actions:
- notify_ae_slack
- write_crm_contact
- pause_sequence
- start_sla_timer_minutes: 30
The three conditions map to the BANT framework: Budget confirmed, Authority (decision-maker on call), and implicitly Need and Timeline from the qualification score. Any AI SDR system worth deploying can evaluate these mid-call and trigger the handoff at the moment all three are true.
The pause_sequence action is critical. The moment a handoff triggers, all automated outreach on that contact stops. Email sequence paused. LinkedIn queue cleared. Any future voice call attempts cancelled. The prospect is now in the hands of a human, and automated touches alongside a live rep engagement is the fastest way to confuse and lose a warm prospect.
Async brief vs live warm intro vs Slack alert: matching format to deal size
Not every handoff deserves the same format. The cost of the format should be proportional to the value of the deal.
| Deal size | Format | Why |
|---|---|---|
| Under £5k | Slack alert + CRM note | Low-cost, rep calls back when free |
| £5k–£25k | Slack alert + automated SMS to prospect | Holds the prospect while rep is briefed |
| Over £25k | Warm transfer (agent stays on line) | Preserves conversation momentum |
| Inbound enquiry | Warm transfer or ≤5 min callback | Prospect is hot; don't let them cool |
The warm transfer is the most technically complex format and the most effective for high-value deals. The voice agent completes the qualification, then says: "Let me connect you now with a member of our team who can talk through the specifics." The agent places the prospect on a brief hold (with music, under 30 seconds), dials the AE, whisper-briefs them in 8 seconds ("Warm lead, Apex Logistics, wants pricing, £15k budget confirmed, decision-maker on line"), and then bridges the two parties.
This format requires the orchestration layer to support three-way call bridging, which Twilio supports natively via <Dial> with conference endpoints. Retell supports warm transfer to a human via its transfer node. The 8-second whisper brief means the AE says hello already knowing who they're talking to and why — the opposite of "I understand you spoke with someone earlier."
We covered the whisper prompt design in detail in our voice agent transfer-to-human guide. For context on how we've built this in a production financial services deployment, see the voice AI and document analysis case study — the handoff architecture there required both warm transfer and async brief modes depending on the call outcome.
CRM write at handoff time: the fields that matter and the fields that don't
The CRM write happens at handoff trigger, not after the AE responds. This matters because the prospect data is freshest at the moment of qualification — by the time an AE manually logs it 2 hours later, they've forgotten the specifics.
Mandatory fields written automatically at handoff:
def write_handoff_to_crm(contact_id: str, call_data: dict):
crm.update_contact(contact_id, {
"lifecycle_stage": "sales_qualified_lead",
"qualification_date": datetime.utcnow().isoformat(),
"qualification_score": call_data["score"],
"call_summary": call_data["ai_summary"],
"intent_signal": call_data["intent"],
"handoff_format": call_data["format"],
"assigned_ae": assign_round_robin(available_aes()),
"sla_deadline": (datetime.utcnow() + timedelta(minutes=30)).isoformat(),
})
crm.create_activity(contact_id, {
"type": "sdr_handoff",
"body": call_data["ai_summary"],
"call_recording_url": call_data["recording_url"],
})
Fields that don't belong in the auto-write (save them for the AE to add manually after the call-back):
- Full call transcript (too long; link to the recording instead)
- Detailed objection breakdown (the call summary covers this adequately)
- Competitor mentions (inferred from transcript; AE confirms on call-back)
- Proposed solution fit (AE assesses this, not the SDR system)
The discipline here is writing what's known at handoff time, not what would be ideal to know. An AI-generated call summary is available immediately; a fully-annotated opportunity note is the AE's job.
SLA on AE pickup: what happens if nobody takes the handoff
The SLA defines the maximum time between handoff trigger and AE contact attempt. Without an SLA, hot leads sit in Slack until someone gets to them.
A practical UK B2B SLA:
handoff_sla:
tier_1_response_minutes: 30 # During business hours
tier_2_escalation_minutes: 60 # Slack alert to sales manager
tier_3_minutes: 240 # Automated holding message to prospect, lead enters nurture
out_of_hours: "Queue for first business hour contact next working day"
The tier-3 action — an automated holding message — is important. If no rep contacts the prospect within 4 hours, the system sends a personalised email or SMS: "Hi [Name], one of our team tried to reach you after your call today — we'll be in touch first thing [next working day]." This prevents the prospect from feeling ghosted while the internal process catches up.
Track SLA compliance weekly. In our experience, SLA breach rate above 15% indicates either an AE capacity problem (need more reps or better scheduling) or a handoff volume problem (qualification criteria too permissive, producing more handoffs than the team can handle).
Measuring handoff quality: conversion from SDR-qualified to AE-booked
The metric that exposes handoff quality is not the reply rate or the qualification rate — it's the conversion from SDR-qualified to AE-booked call.
If that number is below 30% for warm, budget-confirmed leads, the handoff design is the problem, not the AE's pitch. The speed-to-lead guide covers the response-time element in detail — handoff latency and speed-to-lead are the same problem, viewed from different ends. Common patterns:
- 30% → Healthy for most UK B2B use cases
- 15–30% → Investigate latency (SLA compliance) and context (are AEs briefed?)
- Below 15% → Significant handoff design failure; audit the full sequence
Calculate this monthly, segmented by AE (to identify rep-level issues), by handoff format (warm vs async), and by lead source (inbound enquiries should convert higher than cold outbound qualified leads).
What changed in 2025–2026: AI-generated call summaries and real-time transcription
The most significant change in the past 12 months is the quality of automated call summaries. In 2024, AI-generated summaries of qualification calls were reliable enough for internal notes but not good enough to replace a rep's own briefing notes. In 2026, on production deployments we run with Deepgram Nova-3 for transcription and Claude Haiku for summarisation, the generated summaries are accurate enough that AEs are using them as their primary brief for call-backs without adding to or correcting them in most cases.
The summarisation prompt:
Summarise this qualification call in 80 words or fewer. Include:
(1) what the prospect said they need,
(2) budget if mentioned,
(3) timeline if mentioned,
(4) any objections or concerns raised,
(5) what was agreed at the end of the call.
No preamble. Start with the prospect's stated need.
Output from a real (anonymised) deployment: "Prospect needs automated outbound calling for 8 reps currently dialling manually. Budget £12–15k. Wants to go live before end of Q2. Concern: ICO compliance for financial services calls. Agreed to send UK regulatory overview and book a demo for next week."
Eighty words. An AE can read that in 8 seconds before dialling. That's what a good handoff looks like.
Good / Bad / Ugly
Good: Handoff triggers at qualification, pauses all sequences, writes structured CRM data, Slack-notifies the assigned AE with a 5-line brief, starts a 30-minute SLA timer, and sends a warm-hold SMS to the prospect if the AE hasn't responded within 20 minutes. SDR-qualified-to-AE-contacted rate: 89%. SDR-qualified-to-booked rate: 34%.
Bad: Handoff fires a Slack alert with name and number. No CRM write until AE manually logs it. No SLA timer. No sequence pause. Sequences continue running alongside the AE engagement. AE calls back and prospect says "I already got three more emails." Conversion rate: 12%.
Ugly: No handoff automation at all. AE watches a shared spreadsheet. Leads arrive at 2pm, get actioned at 4:30pm. Half the prospects have moved on. The SDR's qualification rate is good. Nobody measures what happens after.
The handoff is not a notification. It's a system. The notification is just the visible part.