What problem does it solve?
Inbound interest comes from people who chose to reach out, and much of it is wasted. Web forms ask for a lot and answer nothing; chat requests go unanswered outside office hours; sales development reps spend their day separating buyers from job seekers and support requests, while the real buyer waits for a reply and may book with a competitor. On the phone, callers navigate menus to reach a salesperson who first has to ask the same questions again.
The deployments on this page show the same pattern in software, car retail and mortgages. At 8x8, more than half of website sessions happened outside business hours, and a meaningful share of chat requests went unanswered, according to its vendor Qualified. A sales team that works office hours in one language leaves the rest of the day and the rest of the market to a form. A form or a scripted chatbot can capture an email address but cannot hold a discovery conversation. Agents that understand the product, ask the next useful question and act in the CRM and calendar are meant to close that gap.
How does it work?
- Engage at the moment of intent. The agent greets the visitor or caller, in their language, with context from the page they are on or the campaign they came from, and says it is an AI.
- Answer first questions from approved content. Product, pricing principles, availability and next steps come from the organization's approved knowledge, with a refusal when it does not know.
- Qualify with discovery questions. It asks what a good salesperson would (need, size, timing, budget, location) and recognises known accounts from the CRM and intent data, instead of presenting a long form.
- Route by rules. Qualification and routing rules decide the next step: book a meeting with the right rep or specialist, transfer a live call, offer self service, or send a support request or job seeker to the right place.
- Write it down. The conversation summary, answers and score go into the CRM so the rep starts where the agent stopped.
- Follow up with consent. Visitors who leave without booking get a follow up by email or messaging only where consent and contact rules allow.
- Audience
- Customer facing
- Autonomy
- Supervised agent
- Adoption
- Early adopters
- Channels
- Web chat, Phone and voice, Email, WhatsApp
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Conversion uplift | Too few to pool | 19% | 1 | 1 vendor |
| Interactions handled | Not pooled | at least 400,000 | 1 | 1 vendor |
Value drivers: Revenue growth, Speed and cycle time, Lower cost to serve, Customer experience.
Indicative value
A B2B software company with 20,000 inbound leads a year
USD 150,000 to USD 1.8 million
Additional first year revenue from better inbound conversion per year
How this is calculated
Formula: leads * baseConversion * uplift * dealValue. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Inbound leads per year leads, leads per year | 20,000 | 20,000 | The reference company. |
| Lead to closed deal conversion today baseConversion, fraction of leads | 0.01 | 0.02 | Editorial assumption for B2B inbound. Replace with your own funnel data. |
| Relative uplift in lead to deal conversion uplift, fraction | 0.05 | 0.15 | Conservative against the evidence on this page (Qualified reports that 8x8 saw 19% better MQL to SQL conversion and 24% more MQL to closed won deals in its first nine months), because vendor case studies select their best results. |
| Average first year deal value dealValue, USD per deal | 15,000 | 30,000 | Editorial assumption. Replace with your own average contract value. |
What it leaves out: Revenue, not margin, and only the conversion effect. It leaves out the cost of the agent, time saved by sales development reps, after hours coverage beyond the leads counted, and the risk that faster qualification also brings forward deals that would have closed anyway.
Who already uses it?
4 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
CarMax
United States · Automotive · 2026
CarMax, the largest used car retailer in the United States, deployed AI voice agents on its inbound sales calls in 2026. The agent asks clarifying questions to understand what the caller needs, answers common questions such as store hours and vehicle availability, and hands the caller to the right associate faster, whatever the call volume or time zone. CarMax reports more calls resolved and fewer unresolved calls without giving figures, and plans appointment management for appraisals, browsing and test drives. It already runs a web virtual assistant, Skye, on CarMax.com.
No outcome disclosed.
8x8
United States · Technology and software · 2025
8x8 had rising inbound traffic but falling meeting volume: chats went unanswered, more than half of website sessions fell outside business hours and sales development reps spent time sorting job seekers and support requests from buyers. It deployed an AI SDR agent that engages visitors around the clock, qualifies them by company size, country and intent, routes meeting and pricing requests and hands ready conversations to sales, with automated email follow up for buyers who did not book. The vendor reports gains across the funnel in the first nine months.
- Conversion uplift: 19%, first nine months live
"+19% MQL → SQL conversion"
Claimed by: vendor - Conversion uplift: 24%, first nine months live
"+24% MQL → closed-won deals"
Claimed by: vendor
Rocket Mortgage
United States · Banking · 2025
Rocket Mortgage runs an AI Digital Assistant across chat and voice that takes prospective borrowers from first questions to preapproval: it answers questions, collects information, pulls credit, presents personalised rates and loan options and hands the client to a human banker. According to the vendor, the programme started as a proof of concept and has grown to more than 400,000 successful chat conversations and over one million outbound dials a month. Clients who start with the assistant close at three times the rate of those who do not. Sierra also reports that clients who use both the AI chat and a banker convert four times better, without stating the comparison group.
- Interactions handled: at least 400,000, successful chat conversations per month
"What started as a proof of concept in May has grown to more than 400,000 successful chat conversations and over one million outbound dials each month, and both are rising fast."
Claimed by: vendor
SUSE
Europe · Technology and software · 2024
SUSE, the open source software company, moved to ungated content and lost the form data that told it who was on its website. It deployed an AI SDR agent that engages every visitor, asks discovery questions adapted to developers, architects or CIOs, recognises returning accounts through Salesforce and intent data, answers in the visitor's language and follows up by email. The vendor reports that the agent turns 70% of qualified conversations into booked meetings and that its email follow up influenced more than USD 10 million in pipeline.
No outcome disclosed.
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Written qualification criteria and routing rules agreed between marketing and sales
- Approved product, pricing principle and competitive content
- CRM account and contact data, and intent data where available
- Consent and contact preference records for follow up
Systems to integrate
- CRM (leads, contacts, accounts, activities)
- Calendar and meeting booking for reps and specialists
- Website, chat and telephony for live transfer
- Marketing automation for consented follow up
- Intent or enrichment data providers
Complexity: Medium
A chat that asks questions is simple. The work is in written qualification rules sales agrees with, CRM and calendar integration with the right routing, product answers that stay within approved claims, and consent handling for follow up.
- 1
Agree the definition of a qualified lead
Write down with sales what makes a lead worth a meeting, who gets which lead, and what happens to the rest. The agent can only be as consistent as the rules.
- 2
Start with after hours and overflow
Put the agent where nobody answers today (nights, weekends, peak campaigns) so the uplift is easy to see and nobody's pipeline is taken away on day one.
- 3
Ground every product answer
Load approved product and pricing content with owners, and make the agent hand over rather than improvise on pricing, discounts, legal terms or roadmap.
- 4
Integrate booking and CRM before launch
A qualified conversation that does not land in the CRM or a rep's calendar is a lost lead. Test routing for every territory and segment.
- 5
Review conversations with sales every week
Sales and marketing read a sample of qualified and disqualified conversations together and adjust questions and rules.
- 6
Measure against a baseline
Compare lead to meeting, meeting to opportunity and closed won rates with the period before, or with a control group, not only the number of conversations.
Guardrails
- No price, discount or contractual commitment unless it comes from a system of record
- Qualification and routing by written rules, with the reason stored in the CRM
- AI disclosure at the start of the conversation and on voice calls
- Follow up only with valid consent and within contact rules and quiet hours
- Protection against prompt injection and attempts to extract confidential information
KPIs to instrument
- Lead to meeting and meeting to opportunity conversion versus baseline
- Speed to first response and share of inbound answered outside business hours
- Share of conversations disqualified and the reasons, with a sample checked by sales
- Closed won revenue from agent qualified leads
- Opt outs and complaints from follow up
Human in the loop
Sales owns the qualification rules and every commercial commitment. Reps take over qualified conversations, and a sales and marketing pair reviews a weekly sample of agent conversations, including disqualified ones, to catch good buyers the rules turned away.
Common failure modes
- The agent makes promises
- Prospects push for prices, discounts or commitments and a fluent agent agrees. Keep commercial terms out of the model and test for manipulation.
- Qualifying out good buyers
- Rigid rules turn away real buyers who answer one question the wrong way. Review disqualified conversations, not only qualified ones.
- A pipeline that sales does not trust
- If rep and agent disagree on what qualified means, reps ignore the leads. Agree the rules first and show the reasoning in the CRM.
- Follow up that breaks consent rules
- Automated email and calls to people who did not consent create regulatory and brand risk. Check consent before every follow up.
What are the risks and rules?
EU AI Act
Limited risk (transparency)
A customer facing sales agent must make clear that people are talking to an AI system, unless that is obvious (Article 50(1)). Qualifying and routing prospects is not an Annex III use. It becomes high risk where the same system takes on an Annex III task, for example evaluating the creditworthiness of natural persons (Annex III point 5(b)) or assessing risk and pricing for life or health insurance (point 5(c)); those decisions then need the high risk controls.
Rules that apply
Guidance
- Article 50, transparency obligations for providers and deployers of certain AI systems (European Union, Europe). Providers must design systems that interact directly with people so that they are informed they are dealing with an AI system, unless this is obvious from the context.
- Direct marketing and privacy and electronic communications (Information Commissioner's Office, Europe). UK rules on consent for marketing emails, texts and calls, which govern automated follow up of leads.
- Declaratory ruling on AI generated voices under the TCPA (FCC 24-17) (Federal Communications Commission, North America). Calls using AI generated voices count as artificial or prerecorded voice calls under the TCPA, so outbound follow up calls need prior consent in the United States.
Controls to put in place
- AI disclosure in chat and at the start of calls
- Consent check before every follow up by email, messaging or phone
- Qualification rules versioned, with the reason for each decision stored
- Audit log of meetings booked and leads routed by the agent
- Regular review of disqualified conversations for bias against segments or regions
When it went wrong elsewhere
- Incident 622: Chevrolet dealer chatbot agrees to sell Tahoe for $1. A car dealer's sales chatbot was manipulated into agreeing to sell a vehicle for one dollar and recommending a competitor, the typical failure of a sales agent without commercial guardrails.
Frequently asked questions
- How much does an AI SDR improve inbound conversion?
- Vendor case studies report gains. Qualified reports that 8x8 saw 19% better MQL to SQL conversion and 24% more MQL to closed won deals in its first nine months with an AI SDR agent. These are vendor selected results with self selection in them; measure your own against a baseline or a control group.
- Should the agent quote prices?
- Only list prices and pricing principles from approved content or a system of record. Discounts and commitments belong to people; the Chevrolet dealer chatbot that "agreed" to a one dollar car shows what happens otherwise.
- Does it work on the phone as well as in chat?
- Yes. CarMax uses AI voice agents on its inbound sales calls to understand what the caller needs, answer common questions such as vehicle availability and pass the caller to the right associate faster, and Rocket Mortgage runs its digital assistant across chat and voice.
How to cite this page
Blits.ai AI Use Case Library, "AI agent for inbound lead qualification and meeting booking", last verified 27 September 2026, https://www.blits.ai/ai-use-cases/inbound-lead-qualification-agent. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 27 September 2026: First published