05 / Products
Funnel/Agentic Sales Intelligence
Find the companies that need you.
Funnel continuously researches your market, identifies evidence of buying intent and turns promising organisations into qualified, evidence-backed opportunities.
Filtration
A wide search space narrows to a few evidence-backed opportunities.
The problem
Most prospecting starts with a list. Teams buy databases containing thousands of organisations and contacts, then try to determine which companies actually matter, which have a current need, what is happening inside each organisation, who owns the problem, what proposition might resonate, and why now is the right time to make contact.
The expensive part of good prospecting is not sending an email. It is understanding who is worth contacting, and why. Funnel automates that research process so your people can concentrate on the conversation.
It is not a bulk email tool. Conventional outbound buys a list, filters by job title and sends at volume. Funnel is built against that model. The expensive part of good prospecting is not sending a message. It is understanding who is worth contacting, and why.
The workflow is simple to state: find the right company, understand why they might buy, and know what to say. Someone on your team decides whether to make contact.
Conventional outbound
How many people can we contact?
- 01
Buy a list
- 02
Filter by job title
- 03
Generate a generic personalised email
- 04
Send at enormous volume
- 05
Hope somebody responds
Funnel
Which people do we actually have a good reason to contact?
- 01
Define the organisations you actually want to work with
- 02
Continuously discover potential customers
- 03
Research each organisation
- 04
Look for evidence of a genuine current need
- 05
Identify the people most relevant to that need
- 06
Determine which proposition is most appropriate
- 07
Explain why this prospect is worth contacting
- 08
Draft an evidence-based approach
- 09
Ask a human to approve it
- 10
Manage appropriate follow-up
- 11
Learn from outcomes
Better research. Fewer messages. Better conversations.
The workflow
Automate the research, not the relationship.
Funnel does the research so your people can concentrate on the conversation. Discovery, qualification and drafting are agentic. Contact remains a human decision.
Figure · From ICP to learning
- 01
ICP
- 02
Discover
- 03
Research
- 04
Detect signals
- 05
Identify buyers
- 06
Qualify
- 07
Match offer
- 08
Draft approach
- 09
Human approval
- 10
Outreach
- 11
Follow up
- 12
Learn
Figure · From a wide search to a short list
Illustrative filtration — not a customer result
- 01
1,000
Organisations in view
- 02
240
ICP matches
- 03
48
Signals
- 04
17
Qualified
- 05
6
High-priority
Most organisations in view never become a conversation. The work is finding the few that should.
Ideal customer profile
Tell Funnel what a good customer looks like.
Funnel starts with an Ideal Customer Profile. That profile is a starting point, not a finished lead list. Matching the filters is not enough. The next step is investigation.
- Geography
- Industry
- Organisation size
- Revenue range
- Employee count
- Technology environment
- Growth stage
- Business model
- Relevant departments
- Strategic priorities
- Technologies in use
- Current initiatives
Discovery
Find the signal before you send the message.
Funnel can continuously discover organisations matching the ICP. It can work across configured public, licensed and customer-connected data sources. We do not publish a vendor integration list the implementation does not yet warrant.
The specific signals depend on your proposition. Growth, recruitment, a new office, funding, a product launch or a published technical problem are useful only where they imply a need you can actually serve.
Possible discovery signals
- Company growth
- Recruitment
- New offices
- Funding
- Acquisitions
- Product launches
- Technology adoption
- Executive appointments
- Organisational change
- Public strategy
- Procurement activity
- Published technical material
Signal detection
Look for a reason to make contact.
Finding a company that matches an ICP is not enough. Funnel investigates whether there is evidence that something is happening now that makes the organisation more likely to need your proposition.
A company recruiting multiple AI engineers may need engineering training. An announced internal AI programme may need architecture. A growing data platform may need analytical technology. A newly appointed executive may be reviewing suppliers. The signal has to match the offer.
Figure · What makes an opportunity
- 01
Company fit
- 02
Buying signal
- 03
Relevant person
- 04
Matched proposition
- Result
Qualified opportunity
Research agents
Specialist investigation, rather than one prompt over a website.
Funnel is easiest to understand as a set of specialist investigations. These are a way of explaining the agentic architecture. They are not a claim that every deployment exposes these roles as named services.
- 01
Company
What the organisation does, its scale, market, structure and current priorities.
- 02
Signal
Change, investment, recruitment, projects, initiatives, problems and evidence of intent.
- 03
Technology
Where relevant: engineering activity, cloud and platform choices, public architecture and hiring requirements.
- 04
Buyer
Which roles are most likely to own the problem, influence the decision or control the budget.
- 05
Qualification
Combine the available evidence into an opportunity the team can rank and review.
- 06
Proposition
Which customer offering best matches the observed need.
- 07
Research
An evidence-backed briefing a person can read before deciding to make contact.
- 08
Writer
A concise personalised approach that uses the research, rather than a greeting plus a job title.
Evidence
Every recommendation should have a reason.
A Funnel opportunity should not simply say that a company looks like a good prospect. It should say why now, what the likely need is, which proposition fits, who to speak to, and what evidence supports that reading. Humans should be able to verify the recommendation.
Illustrative example — not a customer result
Acme Corp
Opportunity score 86
- Why now
- Acme has advertised four AI engineering positions during the last six weeks, including roles referencing retrieval-augmented generation and LLM evaluation.
- Likely need
- The engineering organisation appears to be developing internal generative AI capability.
- Relevant proposition
- AI Engineering for Software Developers
- Recommended buyer
- VP Engineering / Head of AI
Evidence
- Four advertised AI engineering roles in six weeks
- Job descriptions referencing RAG and LLM evaluation
- Public mention of an internal generative AI programme
Opportunity scoring
Rank what is worth a person’s time.
Funnel can rank prospects across several dimensions. The score is a prioritisation mechanism based on configured criteria and available evidence. It is not an objective probability of purchase.
Figure · Prioritisation dimensions
01
86
Fit
How closely the organisation matches the ICP.
02
78
Intent
Whether there is evidence of a current relevant need.
03
72
Timing
Whether there is a reason to approach them now.
04
90
Proposition fit
Whether you have a strong offering for the detected need.
05
68
Contact confidence
Whether an appropriate person or role has been identified.
06
74
Commercial potential
Whether the opportunity appears commercially worthwhile.
Illustrative scores — a prioritisation aid, not a probability of purchase.
Offer matching
Match the offer to the need you observed.
Funnel can understand your product and service catalogue. When it identifies an opportunity, it determines which proposition appears most relevant. Cognitecta uses the same logic on its own pipeline.
AI engineering recruitment
01
AI Engineering training
RAG initiative
02
RAG training or architecture
Agent project
03
Platform
Large analytical data estate
04
Stack
Strategic planning requirement
05
Wargame
Contact-centre transformation
06
Callcenter
AI programme without clear architecture
07
AI discovery / consulting
Custom implementation requirement
08
Build
Personalised outreach
Personalisation should mean understanding the prospect.
Conventional AI email tools often personalise the surface: a first name, a title, a company. Funnel prepares outreach from the observed signal, the organisational context, the likely need, the relevant proposition and the evidence. The result should be concise. Long generated sales letters are not useful.
Illustrative example — not a customer result
Why contact
Acme has advertised four AI engineering positions during the last six weeks, including roles referencing retrieval-augmented generation and LLM evaluation.
Suggested approach
Offer a private RAG / AI engineering programme for the engineering team.
Draft
You are hiring several AI engineers with RAG and evaluation in the brief. We run a private programme for engineering teams who have to put that work into production — not a survey of tools. If useful, we can discuss whether that room would help the people you are hiring now.
Human approval
Research continues until a person on your team is ready to decide.
Funnel should automate discovery, research, qualification, prioritisation, proposition matching, drafting and follow-up administration. Your team remains responsible for high-value relationship decisions: approve, edit, or skip.
Figure · A person approves before anyone is contacted
12 opportunities ready
Acme Corporation
RAG programme detected
87
Northstar Financial
AI platform recruitment
82
Example Industries
Data transformation initiative
78
Review opportunity
Acme Corp
- Why now
- Acme has advertised four AI engineering positions during the last six weeks, including roles referencing retrieval-augmented generation and LLM evaluation.
- Suggested contact
- VP Engineering / Head of AI
- Recommended proposition
- AI Engineering for Software Developers
Interface concept — not a live product screenshot.
Follow-up
Follow up with restraint.
Approved opportunities can enter a follow-up process that is restrained by design: schedule the next step, stop when someone replies, recognise an out-of-office, suppress a decline, and treat “not now” as a later reminder rather than a failure.
- Schedule an appropriate follow-up
- Stop when a prospect responds
- Recognise an out-of-office reply
- Suppress contacts who decline
- Identify not-now opportunities
- Create a later reminder
- Classify the response
- Update opportunity status
Response intelligence
Classify the outcome. Decide the next action.
These workflows should be configurable. The categories are a starting model, not a fixed script.
- 01
Interested
A human takes over the conversation.
- 02
Not now
Return to the pipeline at an appropriate later point.
- 03
Wrong person
Research an appropriate alternative contact.
- 04
Send information
Prepare the relevant material for review.
- 05
Not interested
Suppress further outreach to that contact.
- 06
Out of office
Resume after the stated return date, where one is given.
Learning loop
Learn what actually creates conversations.
Funnel should be able to analyse its own outcomes: which signals correlate with replies, which industries respond, which propositions generate meetings, which roles are effective entry points, which messages fail. That is analytics and feedback that can improve configuration and prioritisation — not a claim that the system rewrites itself.
Figure · Outcomes feed the next configuration
- 01
Discovery
- 02
Outreach
- 03
Response
- 04
Opportunity
- 05
Outcome
- 06
Learning
Dashboard
A sales intelligence view you can inspect.
The useful view is what was researched, which signals appeared, which opportunities are waiting for a person, and why Funnel selected them. Inspect the evidence. Volume on its own is not the achievement.
Figure · A sales intelligence view
Illustrative dashboard — not live customer metrics
248
Companies researched
47
Signals detected
19
Qualified opportunities
8
Awaiting approval
6
Active conversations
Today’s opportunities
| Company | Score | Signal | Proposition | Buyer | Status |
|---|---|---|---|---|---|
| Acme Corporation | 87 | RAG programme | AI Engineering training | VP Engineering | Awaiting approval |
| Northstar Financial | 82 | AI platform hiring | Platform | Head of Architecture | In review |
| Example Industries | 78 | Data transformation | Stack | Director of Analytics | Qualified |
Signal activity
Hiring
34
Technology
22
Investment
16
Leadership
12
Strategy
10
Growth
6
Built to run our own pipeline
We built Funnel for a problem we have ourselves: finding the small number of organisations where there is a genuine reason to start a conversation.
Cognitecta uses Funnel to research its own market, identify organisations investing in AI and surface opportunities for training, consulting and product deployment. It can look for companies hiring AI engineers, launching AI initiatives, investing in RAG or agents, or expanding data infrastructure — then match those opportunities to Training, Consulting, Build, Platform, Stack, Wargame or Callcenter.
That is an internal use, not a published performance claim. We do not invent meetings, conversion rates or revenue attributed to Funnel.
Relationship with Platform
A specialised workflow on agentic foundations.
Funnel applies Cognitecta’s agentic technology to one specific workflow: finding and qualifying B2B opportunities. Platform is the general-purpose agentic foundation. Funnel is a product focused on sales intelligence. That is a statement of family, not a claim about a particular shared runtime in every deployment.
Responsible outreach
Prospecting with controls you can stand behind.
Funnel is designed to support responsible B2B outreach workflows, with human approval, evidence retention and suppression controls forming part of the operating model. Appropriate business contact data, opt-out handling, configurable approval, data governance and auditability belong in that model. We do not make blanket GDPR, PECR or certification claims here.
Capabilities
- 01
Ideal customer profile
Start from the organisations you actually want — geography, industry, scale, stack and priorities — then investigate.
- 02
Continuous discovery
Find organisations that match the profile across configured public, licensed and customer-connected sources.
- 03
Signal detection
Look for evidence that something is happening now that makes a conversation more relevant.
- 04
Buyer identification
Identify the roles most likely to own the problem, influence the decision or control the budget.
- 05
Qualification
Combine fit, intent, timing and proposition into a reviewable opportunity score.
- 06
Offer matching
Choose the proposition that appears most relevant to the observed need, rather than sending one message to everyone.
- 07
Evidence-backed briefing
Preserve the reasons and sources so a person can verify why the opportunity was recommended.
- 08
Drafted approach
Prepare a concise, evidence-based outreach draft. Personalisation means understanding the prospect, rather than greeting them by title.
- 09
Human approval
Automate research up to the relationship. Someone on your team approves, edits or skips before anyone is contacted.
- 10
Controlled follow-up
Schedule appropriate follow-up, stop on a reply, suppress declines and recognise when the answer is not now.
- 11
Response classification
Sort outcomes so the next action is clear: take over, wait, find another person, or stop.
- 12
Learning from outcomes
See which signals, roles and propositions actually create conversations — and use that to refine how Funnel is configured.
Use cases
Where research changes who you contact.
01
Specialist technology sales
The problem
The buyers who need a specialist product are few. A purchased list treats them as a volume problem.
How it is addressed
Define the organisations that can actually use the product, then let Funnel look for the signals that suggest a current need.
02
Professional services
The problem
A useful conversation usually starts from a specific initiative, not from a generic capability brochure.
How it is addressed
Match observed programmes, hiring and published problems to the service that would actually help — then ask someone on your team whether to make contact.
03
Training and capability building
The problem
Teams buy training when they are already building something. Job titles alone do not show that.
How it is addressed
Recruitment, programme announcements and technical publication can be evidence of a team that needs to learn a particular skill now.
04
Product deployment
The problem
A product such as Platform, Stack, Wargame or Callcenter is useful only where the organisational problem already exists.
How it is addressed
Funnel can look for the conditions that make a product relevant — an agent programme, a large analytical estate, a strategic decision under uncertainty, a contact-centre transformation — and surface those organisations first.
Limits worth being clear about
Research at scale. Judgement stays with your team.
- 01
A score is a prioritisation mechanism, not an objective probability that someone will buy.
- 02
Recommendations depend on the quality of the ICP, the available sources and the evidence those sources contain.
- 03
Your team remains responsible for high-value relationship decisions. Funnel drafts; it does not send on its own as a matter of product principle.
- 04
Learning from outcomes is analytics and feedback for configuration. It is not a claim of autonomous self-modification.
Enterprise deployment
These products are delivered as engagements rather than self-serve licences. We start with your problem, your data, and the systems you already run.