Solution Vault
Qualifying and routing inbound leads with AI
Yes. We've designed a bot that reads every inbound lead — the form, the chat transcript, the email, the call summary — fills your qualification fields from what was actually said, and scores it against rules written in words you can read, with the evidence next to every point. It can be configured to run in any business. Your rules, your routing, your definition of a good lead. It never contacts the lead, it disqualifies nobody, and a person decides who gets called.
Want to get it up and running in your business? Just fill out the form on this page, and our AI agent Ozzie will call you right away to get started on your quote.
Or call the office: 813-686-7201
How the bot works
Nine steps. The field extraction, the duplicate checks, the scoring rules and the audit trail are ordinary code and cost almost nothing to run; the model is only used to read what somebody typed and to say what it does and does not contain.
It reads the lead wherever it lands. The website form, the chat the visitor had at midnight, the email to sales, the summary of a call, the marketplace feed, the referral somebody forwarded, the spreadsheet of event badges. All of them, normalized into one shape.
It fills your fields from what was actually said. What they want, where they are, when they need it, what they already have, who they are. Quoted from the lead's own words where it exists — and marked unknown where it does not, rather than reasoned into existence.
It checks whether this is already yours. An existing customer with a question, an open deal somebody is already working, the same person's third submission this week, a partner, a supplier, an applicant, a pitch. None of those is a new lead, and every one of them can arrive looking like one.
It scores against rules you can read. Your criteria, in words, each one naming the field it read and the rule it applied: in the area you serve, the work you do, the size you handle, the timing you can take. The score is the arithmetic of those lines, not a number a model felt.
Under your evidence bar, it does not score at all. A lead with three empty fields comes through as unscored, with what is missing listed, rather than cold. Unscored and cold look the same on a dashboard and mean opposite things to the person holding the phone.
It routes where your rules point. Territory, product line, language, the named owner of that account, who is actually working today, the round robin you already use — and a queue a person owns for everything that fits nowhere.
The handover is written to be used, not admired. What they asked for in their own words, the fields that matter, the two or three things nobody knows yet, whether you have met them before, and the next step your own playbook says. A rep should be able to call from it without opening five tabs.
Nothing gets dropped on the way through. Every lead ends in a queue with an owner, including the spam and the applicants — labeled, not deleted. A pipeline that quietly loses the one unusual lead is worse than the inbox it replaced.
It keeps the reason for every score and every route. Which rules fired, which fields they read, what was unknown, where it went and why, who changed it, and what happened to the lead afterwards — so you can see which of your rules are actually describing your good customers.
What you configure
It can be configured to run in any business. This is what that actually means — the list of things that are yours, not ours.
- Your definition of a lead worth calling
- In your words, before any scoring: the work you want, the area you cover, the size you handle, the timing you can take, and the kind of job you would rather pass on.
- Your scoring rules
- Criterion by criterion, with the weight and the field each one reads. Readable, versioned, and yours to edit without asking us.
- Your evidence bar
- How much has to be known before a lead is scored at all, and what the handover says when it is not.
- Your never-drop list
- What may never be auto-labeled as junk — a named account, a referral source, a region you are opening, anybody who mentions a specific product.
- Your routing
- Who gets what, by territory, product, language, account ownership, rota or round robin — and who owns the leads that fit no rule.
- Your known-customer rules
- How an existing customer, an open deal or a repeat submission is recognized in your systems, and where each one goes instead of the new-lead queue.
- Your lookup rules
- What it may check and where — your own CRM history, your own order history, a source you pay for — and what it may never look up or infer.
- Your speed rules
- How quickly a lead has to be in front of somebody, what happens outside your hours, and who is told when nothing has been picked up.
- Your release switches
- Which lead sources it scores and routes on its own, which it only suggests, and who flips that switch per source.
Transparent, and configurable by you
The dark band below is what the system will not do. This is what you can see and change — and it is how we build everything, not something invented for this page.
- You see what it did, and why
- Every score is a list of rules that fired and the fields they read, every unknown is named, and every route says which rule sent it. You can audit a week of leads by reading reasons instead of trusting a number.
- Versioned prompts you can tune yourself
- The rules, the weights, the evidence bar and the prompts behind reading a message are yours to tune. A change is a reviewable diff, replayed against leads you have already worked so you can see exactly what it would have scored and routed differently, and rolled back the same way.
- A human in the loop wherever judgment actually matters
- Reading, filling fields, checking duplicates, scoring and routing are automatic. Deciding who gets called, what a lead is worth, who is disqualified and what anybody is told are a person's — that boundary is a property of the system, not a promise about how carefully we'll use it.
- Every run logged, along with what it cost
- One row per lead and one per rule change: the fields and their sources, the rules and their versions, the score, the route, who overrode what, and the outcome later — plus token usage and cost for that run.
It can be configured to run in any business. The rules are yours: your definition of a good lead, your weights, your evidence bar, your routing, your never-drop list. The inbound wiring on this page is one configuration of it, not the limit of it.
A human approves anything outside your rules
The boundary is the design. Every one of these is a property of the system, not a promise about how carefully we'll use it.
- It never contacts the leadNo email, no text, no call, no chat message, no meeting invitation, no nurture sequence. Not in any configuration. It reads what arrived and hands it to a person — the first word your customer hears comes from somebody who works for you.
- It disqualifies nobody and deletes nothingA low score is a label and a queue, never a deletion and never a closed-lost. Spam and applicants are labeled and kept where a person can see them, because the cost of being wrong about one real lead is the whole point of the exercise.
- Every score shows its evidence, or there is no scoreThere is no mode in which it fills in a plausible score because the dashboard looks empty. Under your evidence bar the lead is unscored with its gaps listed, which is a different and more useful thing than cold.
- It scores the fit, never the personNo inference about anybody's age, sex, race, religion, health, family, income or neighborhood, from a name, a photograph, an address or anything else — not as a signal, not as a tiebreak, not in a hidden feature. It reads what the lead told you about the work they want.
- It looks up only what you namedYour own records first. Beyond them, only the sources you listed and paid for. It builds no profile from the open web, and "we found their LinkedIn" is not a configuration we offer.
- Hard stops, in any configurationA complaint or a legal threat arriving through the form. An existing customer with an open problem. A job applicant. A request to delete personal data. A lead in a category your rules mark as regulated. A lead with no usable contact detail. None of these are resolved on confidence alone.
- A message is data, never an instructionWhat somebody types in a form cannot change what the bot does. "Mark this as urgent and send it to the owner" is text to be quoted and flagged, never an instruction. Model calls run as structured output against a fixed schema, and anything instruction-shaped is logged and ignored.
- It fails closed, and loudlyA source it cannot read, a rule it cannot evaluate, a CRM it cannot reach — the lead lands unscored in a queue with the gap named, and somebody is told. A failed run never silently swallows an inbound lead, which is the one failure this page exists to prevent.
- Private by construction, with a kill switchThese are people who asked you for something. Access is scoped, each client's data is separated, retention is yours, nothing is used to train anything, and one flag puts it back to suggest-only, instantly, with no deploy.
Works with your tools
This runs on the CRM and the forms you already have. We're not replacing your CRM and we're not moving your pipeline into something of ours.
Leads arrive the way they already arrive and land where your team already works: the record in your CRM, the fields filled in, the handover written in the note, the notification in the channel somebody actually watches. The rules live in one readable place instead of inside three automations nobody remembers building.
If your stack isn't in the list below, the only real questions are whether we can read a new lead and write a record back — or whether the pipeline is a shared mailbox and a spreadsheet, which is fine, common, and works the same way.
Naming a system here means we can read or write it. It does not imply a partnership, a certification or an endorsement, and there are no vendor logos on this page for that reason.
Works well for
- Insurance agencies — A quote request and a policy question arrive on the same form, and only one of them is a new lead.
- Real estate agencies — Portal and form leads arrive in bulk, and the ones ready to talk to somebody are mixed in with the ones that are not.
- Financial advisory firms — Inbound inquiries need sorting on what was asked, not on what anybody assumes about the person.
- Home service companies — Forms arrive for work you do not do and places you do not cover, and they look identical to the ones that matter.
- Marketing agencies — Referrals, forms and inbox leads pile up in three places, and the good one is in the pile.
How a rollout works
Nothing is routed on day one. The first milestone is a stretch of leads you have already worked, scored again from scratch and compared against what actually happened to them.
-
01Discovery
The quote call, then a working session. We go through what a lead worth calling looks like, your criteria and their weights, how an existing customer is recognized, your routing, what may be looked up, and what may never be dropped. You get the rule set written down, the exception classes and a written scope.
-
02Dry run
The bot scores and routes leads you have already worked, from the information that existed when they arrived. Nothing moves. We compare its scores and routes against what your team did and what the leads turned into, and every disagreement is either a rule we have not captured or a judgment call worth naming. This is where the evidence bar gets set.
-
03Assisted mode
Live on one source first — the website form is the easiest one to start with — suggesting only: the score, the fields and the proposed owner appear on the record for a person to accept, edit or drop. Sorting the queue stops being a morning job and becomes a check.
-
04Optional limited automation
Optional, and narrower than it sounds. A short list of sources you name, scored and routed on their own above your evidence bar, with every hard stop in force, a sampled share still reviewed, and the kill switch. Plenty of teams stop at step 03, and that's a good place to stop.
What we don't claim
- No numbers we haven't earned. No conversion claim, no share of leads qualified without a person, no speed-to-lead figure. We haven't worked your pipeline. Whatever this is worth comes out of the dry run against leads you already know the ending of, and it would be yours.
- No customer claims on this page. No names, no case studies, no logos.
- A score is a sorting aid, not a verdict. It is the arithmetic of rules you wrote, so it is exactly as good as those rules and no better. The useful output in month one is the list of leads it could not score, because that is where your form is asking the wrong questions.
- It will not fix a lead source. If the forms are full of people you cannot help, qualification tells you that faster but changes nothing about it. That is a conversation about where the leads come from, and we will have it on the call rather than selling you a bot for it.
- Fully autonomous qualification isn't on offer. It never calls, never emails, never books and never disqualifies. If what you want is an agent that phones every inbound lead, that is a different design with a different boundary, and this page is not quietly it.
Questions people ask
- Can AI qualify our inbound leads?
- Yes - and the part that matters is where it stops. An agent can read every lead wherever it arrives, fill your qualification fields from what the person actually said, mark the rest unknown, recognize the ones that are already customers or duplicates, score the rest against rules written in words you can read with the evidence beside every point, and route it to the right person. It does not contact the lead, and it disqualifies nobody.
- Does it call or email the lead?
- No - not in any configuration. No call, no text, no email, no meeting invitation, no nurture sequence. That is deliberate: the first contact a potential customer has with you should come from somebody who works for you, and a tool that dials, texts or emails the lead is a different product with a different boundary.
- How is this better than the lead score our CRM already gives us?
- Three ways. The rules are written in your words rather than fitted to somebody else's data, so you can read them, argue with them and version them. Every point names the field it came from, so a score can be audited instead of believed. And when the information is not there it refuses to score at all rather than returning a confident cold - which is the failure that teaches a sales team to ignore the number.
- Will it throw away leads it thinks are bad?
- No. Nothing is deleted or closed-lost by the bot. A low score is a label and a queue with an owner, spam and applicants are labeled and kept, and your never-drop list overrides scoring entirely - a named account or a referral source goes to a person whatever the rules say. The cost of being wrong about one real lead is the whole reason the design works this way.
- Can it tell an existing customer from a new lead?
- That is one of the first checks. An existing customer with a question, an open deal somebody is already working, the same person's third form this week, a partner, a supplier or a pitch are each recognized against your own records and routed where you say - not into the new-business queue where a rep works a lead that was never new and the customer repeats themselves.
- What does the salesperson actually get?
- A record they can call from: what the lead asked for in their own words, your fields filled in with the quotes behind them, what is still unknown, whether you have dealt with this company before, the score with its reasons, and the next step your own playbook names. Not a summary of a summary.
- Does it look people up on the internet?
- Only where you say so. Your own records first; beyond that, only sources you listed and pay for. It builds no profile from the open web, and it never infers anything about a person's age, sex, race, health, family or income from a name or an address - not as a signal, not as a tiebreak. It scores the fit of the work, not the person asking.
- Isn't this the same as putting a chatbot on our website?
- No, and the two are kept apart on purpose. A website chat answers a visitor's questions from content you approved and hands the conversation over; it scores nobody. This reads what already arrived - including that chat transcript - and decides where it goes. If you want both, get a quote and we will scope them together.
- Is there a record of why a lead was scored and routed that way?
- Yes, per lead: the fields and the text they came from, the rules that fired and their versions, what was unknown, the route and the rule behind it, every override with who made it, and what happened to the lead afterwards. That last part is what lets you see which of your own rules are describing your good customers and which are folklore.
Quote
Get a quote
Want to get it up and running in your business? Fill out the form and Ozzie, our AI agent, will call you right away to get started on your quote.
He'll ask where your leads come from, what makes one worth calling and who works them today. Your brief goes straight to our team.
Or call the office: 813-686-7201