PetriDigital
Book a call

How B2B AI software actually sells.

Most B2B AI funnels are a channel collection in a trench coat. SEO over here, paid over there, a webinar that was set up in 2024, outbound running unsupervised. This paper is the opposite. Seven stages. Each mapped to a buyer state. Each instrumented with KPIs, tools and the anti-patterns that quietly leak pipeline.

Audit my funnel See the seven stages
01 Demand TOP OF FUNNEL 02 Discovery TOP OF FUNNEL 03 Engage MIDDLE OF FUNNEL 04 Educate MIDDLE OF FUNNEL 05 Validate MIDDLE OF FUNNEL 06 Close BOTTOM OF FUNNEL 07 Compound POST-SALE COMPOUNDING LOOP
The funnel. Seven stages, observed across forty B2B AI software engagements. Stage width tracks surviving volume. Red bars are acquisition. Green bars are engage and post-sale. The dashed arrow is the loop that feeds Stage 07 back to Stage 01.

The seven stages.

  1. 01 Demand Top of funnel
  2. 02 Discovery Top of funnel
  3. 03 Engage Middle of funnel
  4. 04 Educate Middle of funnel
  5. 05 Validate Middle of funnel
  6. 06 Close Bottom of funnel
  7. 07 Compound Post-sale
01

Demand.

The buyer doesn't know you exist. They aren't yet sure the problem you solve is worth budget either.

Buyer state

Problem-unaware or problem-aware. They have a pain (slow analyst workflows, hallucination risk, model ops cost, unstructured data) but no solution category in mind. Their VP hasn't approved a line item.

What we run

Plant the category. We seed thought leadership, executive POVs, and category-defining content where the buyer's inner circle already reads: LinkedIn (organic and paid), PR, podcasts, AI and LLM citations. We don't ask for a meeting. We ask for two minutes of attention.

Anti-patterns

  • Running solution-aware ads at problem-unaware buyers. CTR collapses, CAC bloats.
  • Treating LinkedIn as a billboard instead of a conversation
  • Paying for impressions on lookalikes that don't share intent signals
  • Chasing virality over category authority

Feeds into → Discovery (02). When the buyer remembers the problem next week, they'll search.

02

Discovery.

The buyer searches across Google, ChatGPT, Perplexity, Reddit and G2. They either find you or find three competitors.

Buyer state

Solution-aware. They're typing 'best AI agent platform for ops', or asking ChatGPT 'which RAG vendor handles enterprise security'. They're forming a shortlist in 90 seconds.

What we run

Be the answer wherever buyers research. Ship pillar-and-cluster content built for both Google ranking and LLM ingestion. Earn G2, Reddit and industry-listicle inclusion. Fund high-intent paid to catch the buyer at bottom-of-search before a competitor does.

Anti-patterns

  • Optimising for traffic volume over intent. A million visits to /blog/what-is-rag converts at zero.
  • Skipping schema and HowTo markup so LLMs can't ingest you cleanly
  • Buying bottom-of-funnel paid before the organic foundation can recapture them
  • Trusting agencies that don't track AI citations as a primary KPI

Feeds into → Engage (03). Discovery's job is to deliver the buyer to your owned surface.

03

Engage.

They land on your site. You have eight seconds to prove you understand them better than the next tab.

Buyer state

Vendor-aware. They've narrowed to three to six names. They scan your homepage for: who is this for, what does it actually do, who else uses it, what's it cost, can I trial it.

What we run

Build a website that's fast, accessible, schema-rich, and built to convert. Every page resolves one buyer question. Several low-friction next steps (evergreen webinar, ROI calc, demo, sales call, free trial) sized to where the buyer actually is.

Anti-patterns

  • Hiding pricing without a strategic reason. Most AI buyers leave to a competitor that shows it.
  • WordPress on shared hosting in 2026. Too slow to convert AI buyers.
  • One CTA on the page. Buyers at different stages need different next steps.
  • Auto-playing video on hero. Kills LCP and accessibility.

Feeds into → Educate (04). Engage's job is to win the next 10 minutes of attention.

04

Educate.

They opted in but didn't buy. Now you have weeks to teach them, without going dark and without spamming.

Buyer state

Considering. They've registered, watched the webinar, downloaded the calculator. They're building an internal story for why the category matters, then why you specifically.

What we run

Run an evergreen, multi-touch nurture across email, LinkedIn DM, retargeting and SMS (where consented). The heaviest-lifting asset is the evergreen webinar. It runs 24/7 across timezones and routes high-engagement attendees into the CRM in real time.

Anti-patterns

  • Drip-spamming everyone who downloaded a PDF. It burns the list and kills deliverability.
  • A webinar that's actually a sales pitch. Modern AI buyers leave at minute 7.
  • LinkedIn outbound that pitches before there's any relationship
  • Marketing 'handing off' at MQL. Pipeline stalls when nurture stops.

Feeds into → Validate (05). Educate's job is to graduate buyers into multi-stakeholder buying.

05

Validate.

Your champion now has to convince three people who haven't seen the demo. Most B2B AI funnels die here.

Buyer state

Internally selling. Champion has a budget request, a security review, a procurement form, and a CFO asking 'why this and not GPT', all in parallel. Without ammunition, the deal stalls indefinitely.

What we run

Equip the champion. Multi-stakeholder ABM ads light up the buying committee. Reference customers go on calls. Security trust center, model card, and compliance pack remove the procurement friction. A custom ROI doc lands in the CFO's inbox.

Anti-patterns

  • Single-thread selling. Champion leaves and the deal dies with them.
  • No public security page. Every enterprise buyer asks for one. Missing it kills 30% of deals.
  • Generic ROI calculator. CFOs need their numbers, not a template.
  • Treating ABM as 'expensive paid' instead of 'sales support at scale'

Feeds into → Close (06). Validate's job is to get every stakeholder a 'yes, I can defend this'.

06

Close.

All signals point to buy. Now don't fumble it on pricing, procurement, or 14 days of silence.

Buyer state

Buying. Procurement engaged, legal redlining, security signed, champion calendar-blocked. Deal velocity is the only thing that matters now. Every day idle is a day a competitor calls.

What we run

Smooth the runway. Transparent pricing page where it makes sense. Clean MSAs with pre-negotiated standard redlines. Pre-built procurement-form responses. Deal desk on speed-dial. Executive sponsor available within 24 hours.

Anti-patterns

  • Surprise pricing late in the cycle. Torches trust, often loses the deal.
  • Procurement asks unanswered for over 48 hours. Competitor wins on velocity.
  • Discount-first concession ladder. Race to the bottom.
  • No mutual close plan. Buyer drifts and the deal slips a quarter.

Feeds into → Compound (07). Close's job is to make the customer's first 30 days frictionless.

07

Compound.

The signed customer is the most valuable asset in your funnel, if you treat them like one.

Buyer state

Becoming. They're now an internal champion, a public proof point, a renewal, an expansion deal, an advocate, a referrer, and the next sentence an LLM trains on.

What we run

Wire onboarding to time-to-value, not first login. Run lifecycle marketing post-close. Turn wins into named case studies, video panels and executive testimonial reels. Stand up a customer advisory board. Earn G2 reviews. Publish performance benchmarks. Every output here becomes Stage 01 input.

Anti-patterns

  • Treating customer marketing as 'CS's job'. Kills compounding.
  • Letting marketing disengage at close. Onboarding signal never feeds back.
  • One-and-done case studies. They need refresh and multi-format reuse.
  • Reactive renewals. Should be a programmatic motion.

Feeds into → Demand (01). Every win restarts the funnel louder, better-positioned and more credible than the last cycle.

The funnel math.

Twelve months of funnel for a hypothetical US$3M ARR AI software company on the path to US$10M. Your numbers will differ. The shape will not.

Stage output Volume / yr Conversion
01 Demand impressions 5,000,000 —
02 Discovery sessions on owned surfaces 60,000 1.20%
03 Engaged sessions (2+ pages, 1+ min) 9,000 15.0%
04 Marketing-qualified leads 750 8.33%
05 Sales-accepted leads 180 24.0%
06 Stage-5 opportunities 55 30.5%
07 Closed-won 18 32.7%
08 Active advocates by month 12 11 61.1%
09 Referred opportunities by month 18 16 +45%

Conversion rates differ materially by industry, ICP fit and stack maturity. We model against your actual CRM data in the strategy session, not against the table above.

What we run. What your team runs.

We're a marketing and distribution studio, not a sales agency. Here's where the line sits.

PetriDigital runs

  • Demand. Founder voice, PR, paid social, AI-citation engineering.
  • Discovery. SEO, AI/LLM ranking, paid search, G2 motion.
  • Engage. Website, evergreen webinar, ROI calculator, lead routing.
  • Educate. Lifecycle nurture, LinkedIn outbound, retargeting.
  • Validate. ABM, security trust center, champion enablement, reference programs.
  • Compound. Customer marketing, case studies, G2 review motion, advocate programs.

Your team runs

  • The product. Build the AI platform. We market it.
  • Direct sales. Demos, negotiations, exec sponsor calls.
  • Closing. Pricing, MSA redlines, deal-desk decisions.
  • CS delivery. Onboarding execution, ongoing CSM relationships.
  • Final approvals. Brand voice sign-off, customer story consent, executive interview availability.

Where most B2B AI funnels leak.

Across forty audits, the same five leaks turn up in roughly four of every five funnels. None are exotic. All are fixable inside a quarter.

  1. 01

    No AI-citation strategy.

    Buyers ask ChatGPT, Gemini and Perplexity. If your category is indexed but you aren't named, you're invisible at the moment of highest intent. Most teams aren't even measuring this, let alone optimising for it.

  2. 02

    Engage stage hosted on WordPress.

    Slow load, weak schema, fragile on the long-form, schema-rich content AI search ingests. Discovery delivers the visit. Engage drops the buyer.

  3. 03

    No multi-stakeholder support past Educate.

    Champion has to sell internally with no kit. Deals stall at security review and never restart. ABM, trust center and champion enablement are missing.

  4. 04

    Marketing disengages at close.

    Compound starts the day the contract is signed. Most teams don't even staff it, so the proof, advocates and case studies never come back as Stage 01 input.

  5. 05

    Channels owned by separate vendors.

    SEO at one agency, paid at another, webinars at a freelancer. Nobody owns the connections. The connections are where the compounding lives.

End of paper

Audit your funnel against this paper.

A 15-minute working session with a senior strategist. Bring your CRM dashboard and your three biggest competitors. We'll map your current state against the seven stages and show you the three highest-leverage gaps.

Book the audit → See services