SynthesisLogic

AI agent for lead capture and follow-up

An agent that catches every lead from your website, ads, messages and email, checks the details, scores fit against your rules, gives the lead to the right salesperson within minutes and follows up until the lead answers or says no.

Book an intro call How it works
A morning with this agent
  1. 19 new leads from the website, ads and WhatsAppDone
  2. 3 duplicates merged with existing CRM contactsDone
  3. 7 qualified leads assigned to 3 salespeopleDone
  4. A large lead asks for a call todayAsks a person

An example. Your agent is built around your own work.

How it works

  1. Catch the lead

    Website forms, landing pages, ad lead forms, WhatsApp, email and call-back requests all land in one place. Each lead is saved with its source, the form it came from and what the person agreed to.

  2. Check and qualify

    It validates the email and phone number, looks for a duplicate in the CRM, and asks a few short questions in the chat or by email: what they need, how much, by when. It scores the lead against the rules you wrote and writes down the reason for the score.

  3. Give it to the right person

    By territory, product or round-robin with a cap per salesperson, it assigns the lead, sends an acknowledgement to the lead and tells the salesperson in Slack, Teams or WhatsApp. If nobody touches the lead within the time you set, it is reassigned.

  4. Follow up until an answer

    It creates follow-up tasks at fixed intervals, drafts each message from the last contact, and stops the moment the lead replies, books a meeting or asks not to be contacted. Every touch is written to the CRM.

New to AI agents? How they work, in plain words

Real examples, setup steps and tool connections for this agent

An example conversation

An example of how a conversation runs. The grey lines show what the agent does in your systems meanwhile.

  1. CustomerMorning. What came in overnight?
  2. Read 14 new leads: 6 website forms, 5 Instagram ad forms, 2 WhatsApp enquiries, 1 email. Checked the CRM: 2 already exist as contacts, 1 is a current customer.
  3. AgentEleven new leads. Three were duplicates, and I added their enquiries to the existing records instead of making new ones. Four score high on your rules: budget above 20,000, wanted within two months, in your delivery area.
  4. Assigned the 4 high-scoring leads by territory. Sent each an acknowledgement with the salesperson's name. Posted the 4 summaries to the sales channel.
  5. CustomerWhy is Lina Foods scored high? They only filled in half the form.
  6. AgentThey named a budget of 35,000 and asked for delivery in November, which meets two of your three rules. They left the company size blank, so I asked them in a reply. The reason is saved on the record, next to the form they filled in.
  7. Lina Foods replied with 120 staff. Updated the CRM field. Score unchanged.
  8. CustomerGood. What about the email one?
  9. AgentThat one needs you. The sender wrote that a colleague of yours promised a discount last month. I have not replied or scored it. I put the thread on your task list in the CRM.

How it is built

Every channel feeds one list
Website forms and landing pages post to the agent through a webhook or the form tool's API. Ad lead forms arrive through the ad platform's lead-retrieval API or its export, messages through the official WhatsApp Business Platform, and email through the Gmail API or Microsoft Graph. Each lead is saved with its source and consent record before anything else happens.
Rules score, the language model explains
Your rules decide the score first: size, sector, region, budget, timing. A large language model, the same kind of AI behind ChatGPT and Claude, reads the free-text part of the enquiry and writes a short reason for the score. The reason is stored on the record so a salesperson can check it and disagree.
Enrichment from sources you approve
If you want company details added, the agent reads the lead's own website, public registers and a licensed data provider you choose, and stores the source for every field. We do not scrape social profiles; most networks forbid it and their APIs are partner-only. Contacts are matched by company domain and the email is verified before any message goes out.
Routing and duplicate rules
The CRM is checked for the person and the company before a record is created. Assignment follows your territory, product or round-robin rules with a limit per salesperson, and a lead nobody has touched within your time limit is moved to someone else.
A fixed list of actions
The agent can create or update a lead, assign it, send an acknowledgement, draft or send follow-ups from your approved templates, book a meeting slot and notify the team. It cannot quote a price, promise a discount or mark a deal won.
Approval, limits and consent
A person approves the first messages of every new sequence, and daily sending caps apply. Unsolicited business email is restricted in the EU and the UK, so each message carries an opt-out and the agent checks consent and your suppression list before every send. Opt-outs are written to the CRM at once.
Where it runs, and sensitive records
The system runs in the cloud or on your own servers. Where lead and CRM records are sensitive, the agent, its database, its logs and the language model can all run on your servers, using open models such as Llama, Qwen or Mistral, so the records never reach an outside AI provider. Local open models are smaller than the largest cloud models. For sorting leads, extracting fields and drafting short messages they are usually enough, and we test on your own leads first. Messages sent through WhatsApp, email or ad platforms still pass through those services.

What you can connect to it

HubSpot
Creates contacts and companies, checks for duplicates, assigns owners, and logs every message and outcome on the record.
Salesforce
Reads and writes Lead, Account and Contact records through the REST API, with duplicate checks before each insert.
Pipedrive and other CRMs
Creates persons, organisations and activities through each CRM's API, so the pipeline matches what happened.
Website forms and landing pages
Receives submissions with the form version and consent text, and reads which page and campaign the visitor came from.
Meta, Google and LinkedIn lead forms
Pulls leads from ad lead forms through each platform's lead-retrieval tools or exports, with the campaign recorded on the lead.
WhatsApp Business Platform
Receives enquiries and sends the acknowledgement and approved template follow-ups to people who contacted you.
Gmail and Outlook
Reads enquiry mail and sends follow-ups from the salesperson's own mailbox, checking for a reply first.
Google Calendar and Outlook calendar
Offers open meeting slots to qualified leads and books the one they pick.
Slack or Microsoft Teams
Tells the salesperson about a new lead with its score and reason, and takes questions from the sales manager.
Google Sheets or Excel
Holds the leads and the review queue if you have no CRM, and takes imported lists from trade shows.
Book an intro call

Connects to your tools. You stay in control.

Works with what you have
HubSpot, Salesforce, Pipedrive and other CRMs, website forms, Meta, Google and LinkedIn lead forms, WhatsApp Business Platform, Gmail, Outlook, Google Calendar, Slack or Teams, spreadsheets.
A person approves what matters
Payments, orders, discounts or anything you choose wait for approval. Every action is logged.
Can run on your own servers
The agent, its data, its logs and even the language model (open models such as Llama, Qwen or Mistral) can run on your servers, so documents and customer data stay inside your network. How that works

What it does not do

  • It does not buy lists or email people who never contacted you or agreed to hear from you. Outreach to new prospects, if you want it, starts as drafts a person approves.
  • It does not scrape social network profiles or get around a platform's terms.
  • It does not invent company details. A field with no source stays empty.
  • It does not quote prices, offer discounts or mark a deal won or lost.
  • It does not keep following up after a reply or an opt-out.

Questions

Does it buy lists or send cold spam?

No. It works with leads who contacted you or agreed to hear from you. Outreach to new prospects, if you want it, starts as drafts a person approves, and every message carries an opt-out.

We have no CRM. Is that a problem?

No. The agent can keep leads, scores and the log of touches in a spreadsheet, and we can help you move to a CRM later.

What is the difference between a lead capture form and this?

A form only collects what the person typed. The agent also checks it, finds duplicates, scores it, hands it to a salesperson and keeps following up. The form can stay as it is, since the agent reads what it sends.

How does the lead scoring work?

You write the rules, such as size, sector, region, budget and timing, and the agent applies them. It also writes a short reason for each score and keeps it on the record, so a salesperson can see why a lead was ranked high and correct the rule if it is wrong.

How fast does a lead get a first reply?

Within minutes if you want it. The acknowledgement and the assignment happen as soon as the lead arrives, and the log shows the time to first contact for each lead, which is the number most teams never measure.

Can it find new customers by itself?

It can help, within limits. It can build a list from an ideal customer profile using sources you approve, check each company's public details and draft a first message that a person approves. It does not scrape social networks or send unreviewed mass email.

What if it scores a good lead too low?

Salespeople can override any score, and the override is logged. Each month we compare scores with what actually happened, replies and meetings, and adjust the rules. Anything the agent cannot read well goes to a person unscored.

Can the AI run on our own servers?

Yes. If your lead and CRM data is sensitive, the agent, its database, its logs and the language model can run on your servers with open models such as Llama, Qwen or Mistral. Those models are smaller than the largest cloud ones, but for sorting leads and short drafts they are usually enough, and we test on your own cases first. WhatsApp, email and ad platforms still carry the messages.

How much does an agent like this cost?

It depends on the task and the systems it connects to. After a first call we give you a clear estimate of cost and time.

Want this agent built around your work?

Tell us in a few lines what the job looks like today and we will set up an online intro call.

Type of organisation