Agent Product Page · AI Account Intelligence & Prospecting
NOVA — AI Account Intelligence & Prospecting Engine
WHO TO SELL TO,
AND WHY NOW.
NOVA answers one question every morning: who should we sell to next? It learns what a real customer looks like from your own CRM outcomes, watches the licensed and public signals that say a company is moving, and continuously re-ranks your target account universe by expected revenue — not by lead volume — with the evidence attached to every account.
NOVA value proposition
NOVA answers who to sell to next: a live ICP model, licensed and public buying signals, your own first-party behaviour data, and one ranked list each morning with the reasons attached.
PROBLEMS IT SOLVES
Sales people spending selling time on list building
Lists built on gut feel rather than what actually closes
Duplicate and suppressed contacts leaking into outreach
No record of where a prospect came from when someone asks
Pipeline volume that swings with whoever had time this week
WHO SHOULD WE SELL TO NEXT?
NOVA is not a list builder with a search filter on it. It maintains a live model of who buys from you, watches the public and licensed signals that say a company is moving, and answers one question every morning: of everyone in your market, who should we approach next, and why now?
Analysed continuously
Fit — who looks like a customer
- ICP model learned from closed-won and closed-lost history
- Firmographics: sector, size, structure, geography
- Company growth signals — headcount trend, new locations, expansion
- Published technology stack and tooling changes
- Competitor activity and displacement openings in public sources
Timing — who is moving now
- Hiring signals and job postings that imply a need
- Funding, investment and ownership events
- Trigger events: leadership changes, mergers, new licences, regulatory changes
- Public intent signals from licensed intent providers
- News, filings and announcements relevant to your offer
Engagement — who is already looking
- Website behaviour on your own site, under your own consent policy
- Repeat visits to pricing, service or industry pages
- Content and email engagement history
- Prior conversations, lost deals and dormant accounts worth revisiting
- Referral and existing-customer network overlap
- Learns the profile that actually closes from your own CRM outcomes, not from a guess
- Separates the segments that close fast from the ones that stall or churn
- Refreshes the model on every closed-won and closed-lost result
- Queries approved, licensed data providers and public sources for matching companies
- Identifies the right named entry point at each company
- Deduplicates against your CRM, live pipeline and suppression lists
- Watches hiring, funding, growth, technology, competitor and trigger-event signals
- Weights each signal by how well it predicted wins in your own history
- Combines external signals with behaviour on your own site and in your own records
- Combines fit, timing and engagement into one ranked order
- States the reasons behind every score, in plain language
- Flags low-confidence records for human review rather than pushing them into outreach
01 · ICP intelligence
Who buys from us?
You get: A living ICP with the attributes that predict a win, and what each one is worth.
02 · Discovery
Who matches that profile?
You get: A clean, deduplicated set of in-market companies and contacts, each with its source and licence recorded.
03 · Signals & intent
Who is moving right now?
You get: A timing score per company — what changed, when, where it was observed, and why it matters to you.
04 · Prioritisation
Who should we approach first?
You get: A ranked list with a fit score, a timing score, the evidence for both, and a recommended entry point.
ONE PROSPECT, END TO END
Illustrative example using a fictional company and representative signals — not a client record. On your account every signal would carry its source and licence reference.
- Describe
- [Example Co] moved to the top of this week's list — rank 1 of 214.
- Cause
- Strong ICP fit plus three timing signals inside 30 days.
- Evidence
- Fit: commercial litigation, 24 fee earners — matches the segment behind 6 of your last 10 wins
- Hiring: two open operations roles referencing intake and case management
- Growth: second office announced in the last quarter
- Technology: published stack shows no case-intake automation
- Behaviour: three visits to your litigation page and one pricing view this month
- Prediction
- Companies matching this fit-and-signal pattern have historically been the most responsive segment in your own data; treated as a modelled ranking, not a guaranteed outcome.
- Recommendation
- Approach the Practice Manager now, leading with intake load rather than a generic pitch.
- Confidence
- High — fit score 87/100, timing score 91/100, five independent signals agreeing.
- Expected impact
- Highest expected reply rate of this week's list; queued ahead of 213 lower-ranked records.
- Risk
- Low. No existing CRM record, not on the suppression list, all signals from licensed or public sources.
- Action
- Prepared for HERALD's litigation intake sequence — awaiting your approval before any contact is made.
EVERY ACCOUNT CARRIES ITS OWN SCORES
NOVA does not optimise for list volume. Each account in the universe is scored on expected revenue, not on how easy it was to find, and every score is shown with the evidence behind it.
- ICP fit score
- How closely the account matches the profile behind your actual closed-won deals, learned from CRM outcomes rather than a manual filter.
- Buying intent score
- Weight of observed intent: licensed intent data, content engagement, first-party website activity and competitor movement.
- Timing score
- How recently the trigger events fired — hiring, funding, leadership change, tooling change, advertising shifts.
- Estimated deal value
- A modelled value range from comparable closed-won deals in your own history, used to rank by expected revenue.
- Recommended approach
- The entry point, the angle and the sequence NOVA suggests, based on what has converted for similar accounts.
- Reason for recommendation
- A plain-language explanation listing every signal that moved the ranking, with its source and observation date.
- Confidence level
- How much agreement there is between independent signals — low confidence routes the record to human review instead of outreach.
WHY THIS ACCOUNT?
Illustrative example using a fictional company. Figures and signals are representative, not a client record; on your account each line would carry its source and licence reference.
[Example Ltd] is ranked #3 of 214 accounts in this week's universe.
- Matches 91% of the profile behind your highest-value closed-won deals
- Hired three salespeople in the last 60 days
- Announced an £8m funding round
- Runs HubSpot as its CRM — a stack you integrate with directly
- Increased paid advertising spend, visible in public ad libraries
- Visited your pricing pages twice this month (first-party, consented)
- Has an open Head of Revenue role posted publicly
- ICP fit
- 91 / 100
- Buying intent
- 84 / 100
- Timing
- 88 / 100
- Estimated deal value
- modelled from comparable closed-won deals
- Recommended approach
- Approach the newly appointed revenue lead, leading with ramp time for the new sales hires rather than a generic pitch.
- Confidence
- High — seven independent signals agree, six of them observed in the last 60 days.
NOVA works only from licensed data providers, genuinely public sources, and your own first-party records collected under your consent policy. It does not scrape gated, private or password-protected sources, and it never bypasses a paywall or a site's access rules. Every signal carries its source and the date it was observed, scores are modelled rankings rather than predictions of a sale, and no contact is made until a human approves the list.
INPUTS, LOGIC, OUTPUTS
Inputs
- Closed-won and closed-lost CRM history
- Your ICP rules, exclusions and target geographies
- Licensed B2B data and enrichment provider feeds
- Existing CRM records, open pipeline and suppression lists
- Public buying signals (hiring, funding, tech, expansion)
Processing Logic
- Derive an ICP model from the deals that actually closed
- Query approved providers for companies and named contacts matching it
- Enrich, normalise and validate each record
- Match against CRM, open pipeline and suppression lists to remove duplicates
- Score 0–100 for fit and attach the reasons behind the score
- Rank, batch and post the list for human review
Outputs
- A ranked prospect list with fit scores and score reasons
- Enriched CRM records with source and licence reference
- A refreshed ICP definition after each closed-won deal
- An approval queue for anything below your confidence threshold
- A handover payload for HERALD
NOVA specification
Connects To
- HubSpot, Salesforce or Pipedrive
- Licensed B2B data and enrichment providers
- LinkedIn Sales Navigator exports
- Your suppression / do-not-contact lists
- Slack or email for daily list summaries
- HERALD (outreach) and CORTEX (analytics)
What Triggers It
- Scheduled run (typically nightly or weekly)
- A new ICP segment added by your team
- Pipeline falling below a target volume you set
- A closed-won deal, which refreshes the ICP model
- Manual run requested from the dashboard
Decisions It Makes
- Whether a company matches the ICP, and with what confidence
- Which contact at that company is the right entry point
- Which records to discard as duplicates, suppressed or out of scope
- Ranking order — who gets contacted first
- Whether a record needs human review before outreach
Acts Autonomously On
- Run scheduled sourcing and enrichment
- Deduplicate and suppress records
- Score and rank prospects
- Create draft CRM records in a pending state
- Refresh the ICP model from new closed-won data
GUARDRAILS, IN WRITING
An autonomous agent is only useful if its limits are explicit. Here is what NOVA will never do, and where a human always signs off.
What It Cannot Do
- Scrape private, gated or password-protected sources
- Contact anyone on a suppression or do-not-contact list
- Buy data outside the providers you have approved
- Sign contracts, agree discounts or commit your business to anything
- Invent facts, case outcomes, guarantees or performance claims
- Give legal, medical or regulated financial advice
- Act outside the systems and data scopes you explicitly grant it
Human Approval Points
- New ICP definitions and segment expansions are signed off by you
- The first list from any new segment is human-reviewed before outreach
- Any prospect scored below your confidence threshold is queued for review
- Data-source changes require explicit approval
WHEN THINGS GO WRONG
Failure Handling
- Every run is idempotent — a retry can never duplicate a record or a send
- Transient failures retry with backoff; persistent failures stop the run rather than guess
- A failed step never silently continues — the workflow halts in a known state
- Failures raise an alert to the owning human with the run ID and the exact failing step
- Errors, retries and halts are written to the same audit trail as successful actions
- If a data provider is unavailable, the run pauses rather than substituting an unapproved source
Escalation To Humans
- Any new segment’s first list goes to a human before outreach
- Records below your confidence threshold are queued, never auto-sent
- Ambiguous entity matches are escalated rather than guessed
- Provider or licence changes stop the run and notify the owner
EXAMPLE OUTPUT
Prospect record written to CRM — a representative example of what NOVA writes back into your systems.
company : [Example Co] — commercial litigation, 24 fee earners
contact : Practice Manager (named contact, business email)
fit_score : 87 / 100
score_reasons : sector match, headcount band match, 2 open vacancies
source : approved B2B provider · licence ref on file
dedupe : no existing CRM record, not on suppression list
next_action : queued → HERALD sequence 'Litigation intake'
human_review : approved by A. Operator, 09:14 UTCEXAMPLE DASHBOARD
NOVA — sourcing dashboard. Every figure below is a simulated demo value, not a reported client result.
Prospects sourced
1,240
last 30 days · illustrative
Passed ICP threshold
38%
score ≥ 70
Duplicates suppressed
206
already in CRM or opted out
Awaiting human review
41
below confidence threshold
Handed to HERALD
432
approved and sequenced
Avg fit score
74 / 100
with reasons attached
KPIs NOVA IS HELD TO
ICP match rate
Share of sourced records passing your fit threshold
Duplicate rate
Records rejected as already in CRM or suppressed
List-to-reply rate
Replies generated per hundred prospects handed to HERALD
Cost per qualified prospect
Sourcing and enrichment cost ÷ approved prospects
Research hours reclaimed
Manual list-building hours removed per month
Technical Architecture
HOW NOVA CONNECTS
NOVA reads from your existing systems, applies its own governed processing, and writes results back. Nothing runs outside these lanes.
Upstream · sources
- CRM closed-won / closed-lost history
- Licensed B2B data + enrichment providers
- Suppression and do-not-contact lists
- Public buying signals
NOVA · processing
- ICP model builder
- Sourcing + enrichment pipeline
- Dedupe and suppression matcher
- Fit scoring engine (0–100, reasons attached)
- Human review queue
Downstream · destinations
- CRM prospect records
- HERALD outreach queue
- CORTEX analytics feed
- Slack / email list summary
- Runs on schedule inside Xiilio infrastructure — no agent has standing write access to your CRM beyond the scoped connection
- Every sourced record carries provider and licence reference through to the audit log
SECURITY & AUDIT LOGS
Security
- Data encrypted in transit (TLS 1.2+) and at rest
- Least-privilege access — the agent only sees the records its job needs
- Row-level security on every customer record in the Xiilio platform
- UK/EU data handling with GDPR rights (access, export, deletion) honoured
- No customer data used to train third-party foundation models
Logged On Every Action
- Timestamp (UTC) and agent identity
- Trigger that started the run
- Every system read and write, with the record ID touched
- The decision taken and the rule or score behind it
- Human approvals, overrides and rejections
- Data provider and licence reference for every sourced record
Full detail on encryption, data handling and your GDPR rights is on the Trust & Security page.
How a litigation firm would use NOVA
A commercial litigation practice wants more mid-size corporate clients. NOVA reads their last two years of instructed matters, learns the profile that actually converts, and builds a weekly ranked list of matching companies with a named contact each. The practice manager reviews the list in minutes instead of building it over days, then approves it straight into outreach.
This scenario is written to show how NOVA is deployed. It is not a named client, a delivered engagement or a reported result.
Live Demo
SEE NOVA RUNNING
The homepage system builder shows NOVA's discovery stage running inside a full pipeline. Values shown there are simulated.
Build a system with NOVA in itNOVA FAQs
NOVA — QUESTIONS WE GET ASKED
- Where does NOVA get its data?
- From licensed B2B data and enrichment providers you approve, plus your own CRM. It does not scrape private, gated or password-protected sources, and every sourced record carries its provider and licence reference in the audit log.
- How does NOVA decide who is a good prospect?
- It builds an ideal customer profile from your closed-won and closed-lost history — sector, size, geography, role and deal shape — then scores each prospect 0–100 against it. The reasons behind each score are attached to the record so you can check the logic.
- Can we review the list before anyone is contacted?
- Yes. The first list from any new segment is human-reviewed by default, and you can keep manual approval on permanently. Prospects scoring below your confidence threshold are always queued for review.
- How does NOVA avoid contacting people we shouldn't?
- Every record is checked against your CRM, your suppression and do-not-contact lists, and existing open pipeline before it can be passed to outreach.