Outbound flow: software for Amazon and ecommerce sellers
From public research to a lead who has our report. Worked example: Helium 10.
Step 1 · Task
Find two kinds of companies in one vertical: triggers that visibly run their own operations on AI, to name as proof, and targets that should buy, with AI in the product, little AI in operations and money behind them.
Inputs
- Vertical: software companies that serve Amazon and ecommerce sellers
- The L0–L5 maturity scale, scored separately for AI in the product and AI in operations
- Trigger bar: two or more signal types from the last 18 months, at least one with a number (AI share of own support, an all-staff AI mandate, AI in non-engineering job ads, flat headcount while growing, internal AI-ops roles)
- Target bar: PE or VC money, roughly $25–150M revenue, no evidence of AI running any internal function
- Public sources: company news, job posts, SEC filings and earnings calls, funding announcements; revenue and headcount estimates from GetLatka, Sacra, Growjo, Revelio Labs and PitchBook
Expected result
- A shortlist of triggers, each with dated, linked evidence and an operations level
- Targets ranked by fit, with owner or investors, revenue, staff and the visible pain
- Which trigger to name to which target: one in-category company for relevance, one peer with numbers for proof
- Companies left out, and why
What we found
Examples to name: Omnisend is the best trigger. It is a 246-person ecommerce SaaS that Shopify sellers know. AI handles about 40% of its own support tickets, finance automated up to 85% of invoice processing, and it has a Head of AI Operations. VTEX has the hardest evidence: SEC filings show its support staff cut by more than half while revenue grew.
Who to sell to: Helium 10, Threecolts, Teikametrics and Quartile are the strongest targets. Each has PE or VC money behind it, an estimated $45–105M in revenue, and little or no AI in how it runs.
Examples: companies already running operations on AI
Seven companies clear the bar; the first four have numbers on their own support or headcount. Levels are for AI in operations, on the L0–L5 scale.
1. Omnisend
Ops L3–L4PeerEmail and SMS for Shopify sellers · ~246 staff · ~$50M ARR
2. VTEX
Ops L3–L4PeerEnterprise commerce platform · NYSE · ~1,140 staff
3. Yotpo
Ops L3PeerReviews and loyalty for ecommerce · ~600 staff
4. Hostinger
Ops L3PeerWebsite and store builder for SMBs · ~900 staff
5. Gelato
Ops L3–L4Peer · one press sourcePrint-on-demand for Shopify and Etsy sellers · ~150 staff
6. Skai
Ops L2–L2.5In categoryRetail media and Amazon ads, Pacvue's rival · ~400 staff
7. Pattern
Ops L2In categoryAmazon seller accelerator · NASDAQ · ~2,300 staff
Category anchors, not peers
Targets: who to sell to, and proof they can pay
Software companies serving Amazon and ecommerce sellers, with money behind them and little AI in their own operations. Ordered by fit. Revenue is a third-party estimate unless marked; none of these companies discloses it.
Helium 10 (Pacvue)
Strong fitOps L1.5- Owner
- Pacvue (formerly Assembly). PSG invested in 2020; Advent led a 2021 round at a $1B+ valuation
- Revenue
- ~$104M ARR for Pacvue group, 2024; ~$21M for Helium 10 alone (likely low)
- Staff
- ~190 at Helium 10; ~407 at Pacvue, Mar 2026, growing
- Pain
- Trustpilot 2.1 from 712 reviews: plan changes, billing, refunds (see step 2)
Threecolts
Strong fitOps L2- Investors
- $90M Series A, Mar 2023 (CoVenture, Crossbeam, General Global Capital, Stratos)
- Revenue
- ~$67M, Sep 2025
- Staff
- ~423, Mar 2026, down from 457
- Pain
- Roll-up of ~24 acquired tools run as one company; billing disputes across brands (see step 2)
Teikametrics
Strong fitOps L1- Investors
- $40M Series B, Jul 2021, led by Intel Capital; ~$65M raised in total
- Revenue
- ~$45M, Oct 2024
- Staff
- ~236, Dec 2024; down ~8% over 12 months
- Gap
- Only one posting mentions AI; nothing in support or ops
Quartile
Strong fitOps L0–L1- Owner
- Rockbridge Growth Equity (PE), 2020; merged with Sidecar, Dec 2021
- Revenue
- ~$51M, Jul 2025
- Staff
- ~286 to ~463 by source
- Gap
- Ads software plus a large service team; no internal AI signal
Jungle Scout
Fit, revenue borderlineOps L2- Owner
- Summit Partners, $110M, Mar 2021; reported majority owner
- Revenue
- ~$20M, Dec 2025, down from ~$50M in 2021
- Staff
- ~180, Dec 2025, down from 218
- Pain
- Shrinking team; long-time customers pushed onto pricier plans (see step 2). A PE owner under margin pressure is a reason to buy now
Gorgias
Fit, may build in-houseOps L2- Investors
- ~$104M raised; $29M Series C-2, May 2024, at ~$530M (Sapphire, CRV, Shopify)
- Revenue
- ~$69M ARR, 2024 (Sacra)
- Staff
- ~416, Mar 2026
- Gap
- Hiring for an AI-native CX function, so it may do the work itself
Carbon6
Fit, ownership unclearOps L1–L2- Owner
- SPS Commerce (NASDAQ: SPSC), bought for ~$210M; closed Feb 2025
- Revenue
- ~$40M expected for FY2025 (SPS forecast)
- Staff
- ~169
- Watch
- SPS lost 350 Amazon seller customers in Q4 2025 and sold the seller revenue-recovery business at a loss in Jun 2026. Check who now owns which Carbon6 brands before outreach
Perpetua
Weak fitOps L0–L1- Owner
- Bought by Ascential in 2021 for $52M plus up to $162M; now under Omnicom via Flywheel
- Revenue
- No reliable figure
- Staff
- ~249, 2023
- Why weak
- Decisions sit inside a global agency holding
Feedvisor
Weak fitOps L0–L1- Investors
- $33M raised; last round 2017 (General Catalyst)
- Revenue
- ~$12M (stale estimate)
- Staff
- ~83, Oct 2024
- Why weak
- Below the $25M line, no new money since 2017
Seller Suite (Worldeye)
UnknownOps L0–L0.5Too small for the ICP
ZonGuru, AMZScout, SmartScout, SellerSprite and Sellerboard are founder-run or bootstrapped, with roughly 15–80 staff and no outside money found. They need AI in operations as much as anyone, but they are unlikely to pay for a consultancy.
Who to name to whom
Name one in-category company for relevance and one peer with numbers for proof.
Helium 10 (Pacvue)
Name Skai and Omnisend
"Skai, Pacvue's closest rival in retail media, restructured in June around AI and hired a Director of AI Enablement. Omnisend, at 250 people, now resolves about 40% of its own support tickets with AI. Helium 10 ships an agent to sellers; its own support and billing queue doesn't run on one yet."
Threecolts, Teikametrics, Quartile
Name Pattern and Omnisend
"Pattern asks its sales reps to use Claude daily and has a corporate AI team building tools for non-technical staff. Omnisend resolves about 40% of its own support tickets with AI and automated up to 85% of invoice processing."
Jungle Scout
Name Pattern and Omnisend
"Pattern now asks its sales reps to use Claude daily and has a corporate AI team building tools for non-technical staff. Omnisend automated up to 85% of invoice processing in one quarter."
Carbon6, Seller Suite
Name VTEX
"VTEX cut its support team by more than half while revenue grew, and its CFO credits AI support automation for the margin gain."
Keep the wording to what the source says: "restructured around AI", not "replaced staff with AI", unless the source says so.
Criteria and sources
- Example bar: two or more signal types from the last 18 months, at least one with a number: AI share of own support, an all-staff AI mandate, AI in non-engineering job ads, flat or falling headcount while growing, executives quoting internal numbers, internal AI-ops roles.
- Target bar: PE or VC money, roughly $25–150M revenue, and no evidence of AI running any internal function.
- Money sources: deal announcements and filings where they exist; otherwise estimates from GetLatka, Sacra, Growjo and Revelio Labs. Estimates disagree, so treat them as a first filter and confirm in PitchBook.
- Not used: Klarna (CEO walked back AI-only support), Etsy (CEO said its cut was not driven by AI), Fiverr (revenue fell 10% after its AI-first cut).
- Source quality: every link was opened. Omnisend's support figure is the CEO's own claim in an interview; Gelato rests on one Swedish press article; Yotpo's numbers are 14 months old; VTEX's filing reports the support cut but does not itself attribute it to AI.
Step 2 · Task
For each target, find what hurts its customers and its staff, and mark every problem: AI can fix it, AI can partly fix it, or it is not an AI problem.
Inputs
- The targets from step 1
- Customer reviews: Trustpilot, G2, Capterra and the Shopify App Store, 1–3★, 2025–2026
- Staff reviews from Glassdoor, Indeed and RepVue, and open job posts that describe work done by hand
- Case studies and trigger evidence, to show who already fixed each problem with AI
Expected result
- Problems counted per target, each with links to the evidence
- An AI-fit mark for every problem; the ones AI does not fix stay listed as context
- For each fixable problem, companies that already fixed it with AI, with a number
- Sample sizes and limits stated, so counts read as direction, not rates
What we found
We found 14 problems: 7 that customers complain about and 7 that employees describe. Nine can be fixed with AI automation. The rest come from company policy, bugs or management decisions, so they are listed for context but not for the offer.
The internal problems fit the offer best. Every target describes overloaded teams after layoffs, and most are hiring people to build client reports, reconcile payments or answer tickets by hand.
All problems at a glance
Seen at = how many of the seven targets show the problem in our samples.
Internal problems: what employees and job posts show
Counts are staff reviews mentioning the problem, from Glassdoor, Indeed and RepVue (2024–2026). The last row comes from open job posts. Darker cells mean a larger share of that company's sample.
| Problem | Helium 10 / Pacvuen=31 | Gorgiasn=34 | Jungle Scoutn=25 | Threecoltsn=24 | Teikametricsn=26 | Quartilen=26 | SPS (Carbon6)n=19 |
|---|---|---|---|---|---|---|---|
| Overload, long hours AI | 6 | 12 | 2 | – | 9 | 6 | 4 |
| Layoffs: same work, fewer people context | 9 | 6 | 5 | 3 | 3 | – | 9* |
| Turnover, knowledge loss AI | 5 | – | 4 | 3 | 4 | 1 | – |
| Weak training, slow ramp AI | 4 | 3 | 2 | – | 1 | 4 | 2 |
| Messy tools, processes, no SOPs partly | 4 | 5 | – | 3 | 5 | 6 | 4 |
| Sales: missed quota, weak leads partly | – | – | 4 · 27.5% at quota | 2 | 3 · 14% at quota | 6 | – |
| Shifting priorities not AI | 7 | 13 | 8 | 5 | – | – | 4 |
| Manual work in open roles | ~10 account managers building Excel reports | Billing support at 1k tickets/mo, 4k expected | CSMs with 40–60 accounts each | AR across 5+ payment systems | Account managers preparing business reviews | CSMs with 15+ accounts, QBRs in Excel | AR ops and deductions (titles only) |
| Glassdoor rating | 2.9 / 3.3 | 3.5 | 3.7 | 3.6 | 3.5 | 4.1 | 3.7 |
Overloaded teams after layoffs
AI can fixAll seven · worst at Gorgias, Teikametrics and Quartile. AI cannot undo a layoff, but it can take the repetitive volume the remaining people now carry.
Evidence
- Gorgias: "Burnout daily" and an understaffed team (Sep–Oct 2026); support associate on 12-hour days (Jan 2026)
- Helium 10: 28 project and product owners cut in one day, work spread across the rest (Mar 2025)
- Teikametrics: remaining staff cover the work of 1–3 departed colleagues (Aug 2025)
- Quartile: customer success often works 12-hour days (May 2025)
- SPS (Carbon6's owner): outdated support structure focused on closing tickets (Nov 2025)
Who fixed it with AI
- Hostinger: 91% of ~1.5M monthly conversations close without a person; specialists now coach the AI on hard cases (Sep 2026)
- Synthesia: monthly support demand grew from 40k to 316k people with no added headcount; 55% resolved by AI
- Clay: 3 support staff handle ~7k tickets a month with 45–50% resolved by AI
Account managers building client reports by hand
AI can fixManaged-service teams · Pacvue, Quartile, Teikametrics, Jungle Scout
Evidence
- Pacvue: Retail Media Account Manager builds weekly and monthly client reports; Data Account Manager works mainly in Excel and Sheets (about 10 such roles open)
- Quartile: Customer Strategy Manager with 15+ accounts, reports and QBRs in Excel; staff ask for fewer accounts per CSM (May 2025)
- Teikametrics: Associate CSM prepares recurring business reviews; India-based AMs working until midnight on US hours
- Jungle Scout: CSM with 40–60 accounts keeps health scores, renewal forecasts and CRM current (Jul 2026)
Who fixed it with AI
- AMZ Advisers (Amazon agency, 85+ accounts): AI drafts monthly client reports; 2–3 hours down to 30–40 minutes each, 120–150 hours a month saved (Apr 2026)
- Storable: CSM account audit from 3–4 hours to about 10 minutes
- Abnormal: QBR prep from a couple of hours to 5–10 minutes
Knowledge leaves with people; new hires ramp slowly
AI can fixAll seven · worst at Jungle Scout, Quartile, Helium 10 and Teikametrics
Evidence
- Jungle Scout: 19 reps came and went since Jan 2025 (May 2026)
- Quartile: many account managers hired at once with little training (Sep 2024); needs SOPs (Indeed, Feb 2024)
- Helium 10 / Pacvue: turnover and sink-or-swim onboarding (2024–2026)
- Teikametrics: no systematic onboarding (Jun 2024)
- SPS: wrong or missing documentation (Oct 2025)
Who fixed it with AI
Billing tickets, collections and reconciliation by hand
AI can fixGorgias, Threecolts, Pacvue, Carbon6 · the internal side of billing. The refund policy itself is not AI's to fix (see below), but the manual work around it is.
Evidence
- Gorgias: billing support handles ~1k tickets a month, expected to reach 4k; chargebacks and invoice audits
- Threecolts: accountant invoices across several billing systems and reconciles 5+ payment gateways monthly
- Pacvue: commission withheld while client payments go uncollected (Nov 2024); sales reps chase invoices themselves (Jan 2024)
- Carbon6: hiring an AR Operations Manager (title only)
Who fixed it with AI
Sales reps researching and following up by hand
AI can fixJungle Scout, Teikametrics, Quartile · AI fixes research and follow-up; lead quality and quota setting are management's job
Evidence
- Jungle Scout: 27.5% of reps at quota, inbound lead flow 2.8 of 5 (RepVue); AEs doing manual cold outreach (2026)
- Teikametrics: ~14% of reps at quota, inbound lead flow 1.8, training 2.2 (RepVue)
- Quartile: SDR role: cold calls, emails and Salesforce updates by hand (open)
Who fixed it with AI
Scattered tools and data
PartlyThreecolts, Pacvue, Gorgias, SPS · AI can search and answer across Slack, docs and CRM. It cannot choose the tools or buy CRM seats.
Evidence
- Threecolts: customer data in out-of-date spreadsheets, no extra Salesforce seats, a weekly manual report (2026); hiring for mass CRM data cleanup
- Gorgias: features shipped without telling support
- SPS: too many tools, redundant routine tasks (Oct 2025)
Who fixed it with AI
Customer problems: what customers complain about
Complaints per target, counted in recent negative reviews. One review can count in several rows.
| Problem | Helium 10n=50 | Threecoltsn=23 | Teikametricsn=17 | Gorgiasn=40 | Quartilesmall n | Jungle Scoutn=12 | Carbon6n=7 |
|---|---|---|---|---|---|---|---|
| Slow or no support reply AI | 12 | 8 | 2 + 3 G2 | 17 | – | 2 | 4 |
| Bot-only or canned support AI | 8 | – | 2 | 5 | – | 1 | – |
| Onboarding confusing or thin AI | 5 | – | 2 | 7 | 3 | 2 | – |
| Account manager turnover or silence AI | – | – | 2 | 3 | 7 | – | – |
| Negative reviews answered AI | none since Apr 2025 | – | none | weeks late | – | 12% | – |
| Charged after cancel or without warning not AI | 18 | 9 | 5 | 7 | 3 | 2 | – |
| Refund refused or slow not AI | 15 | 8 | 4 | 3 | 2 | 2 | – |
| Plan, price or fee change not AI | 19 | – | 3 | 6 | – | 3 | 2 |
| Quality dropped after acquisition not AI | – | 4 | – | – | – | – | 1 |
Slow support, and bots that loop
AI can fixSix of seven · Gorgias worst (17 of 40). Several targets already use a bot, and customers complain about it, so the fix is an agent that resolves, not one that deflects.
Evidence
- Helium 10: about 20 emails without a fix (Mar 2026); AI-only replies (Jul 2026); 8+ requests just to cancel (Apr 2026)
- Gorgias: 4 emails in 3 days, no answer (Feb 2026); bot said the ticket was open for months (May 2026)
- Threecolts: email and WhatsApp unanswered (Jul 2025)
- Teikametrics: AI replies signed as a person (Nov 2025)
- Carbon6 (Seller Investigators): no replies, missed claim deadlines (Oct 2026)
Who fixed it with AI
- Lightspeed: CFO says AI resolves more than 80% of support tickets (May 2026)
- Intercom: 81% of its own support resolved by AI; about 100 support hires avoided (Mar 2026)
- Omnisend: ~40% of tickets, satisfaction now level with the human team after two people trained the agent (Sep 2026)
Onboarding that depends on who you get
AI can fixFive of seven · Gorgias most (7 of 40)
Evidence
- Gorgias: no onboarding, handed to a consultant, customer left (Apr 2026)
- Helium 10: no onboarding contact (Apr 2025)
- Teikametrics: disorganized agency onboarding (May 2025)
- Quartile: rocky first one to two months (G2, Nov 2025)
Who fixed it with AI
- Performio: AI implementation agents cut a customer rollout from 20 weeks to 3.5 (Jun 2026)
- Gelato: carrier onboarding from 5 days to 10 minutes
- SPS Commerce: simple onboarding from days to minutes (Jul 2026)
Account managers who change or go quiet
AI can fixManaged-service targets · Quartile most (7). The customer-side view of internal problems 2 and 3.
Evidence
- Quartile: account manager changed twice in a year (G2, Jun 2026); two meetings missed, emails ignored (Feb 2026)
- Teikametrics: 10-year client, second manager left, weeks of silence (Oct 2025)
- Gorgias: dedicated CSM removed after 8 years (Nov 2025)
Who fixed it with AI
Negative reviews left unanswered
AI can fixHelium 10, Teikametrics, Gorgias, Jungle Scout · cheap to fix: drafting replies for approval
Evidence
- Helium 10: no reply to any of the 45 negative reviews since Apr 2025
- Teikametrics: Trustpilot notes it has not replied to negative reviews
- Gorgias: replies on the Shopify App Store came about two months after the review
- Jungle Scout: replied to 1 of 7 negative reviews since Mar 2025
Who fixed it with AI
- Toca Boca: reply rate from 36% to 75% with automation and AI-rewritten templates
- Stellar Service Brands: reply rate from 60% to 87% across 400 locations (neither is a SaaS company)
Not an AI problem: listed for context
These come from billing rules, pricing decisions, bugs or management. They are real and frequent, but AI automation cannot fix them. Where a small part is operational, it is noted.
- Charged after cancelling, refund refused. The most common complaint: Helium 10 (18 of 50), Threecolts (9 of 23), Teikametrics (5 of 17), Gorgias, Quartile. Billing rules or bugs. Operational part: cancellation requests stuck in the support queue belong to "Slow support".
- Plan, price and fee changes. Helium 10's single biggest complaint (19 of 50), after the Platinum cuts announced by email only and a non-refundable 2% ad-spend fee; also Jungle Scout, Teikametrics, Carbon6. Company policy.
- Quality drop after acquisitions. Threecolts customers date problems to the takeover; SPS lost 350 seller customers in a quarter while integrating Carbon6. Integration decisions. Operational part: one help desk across brands belongs to "Scattered tools and data".
- Shifting priorities and leadership churn. The most common staff complaint at Gorgias (13 of 34) and Jungle Scout (8 of 25). Management.
How the triggers run the same work
The difference shows in their job posts: triggers hire people to own and coach the AI, while targets hire people to do the work by hand.
| Trigger | Role | How AI is written into it |
|---|---|---|
| Omnisend | Customer Success & AI Operations Manager (May 2026) | Decides what is AI-owned, AI-assisted or human-owned; measured on coverage and manual work removed |
| Hostinger | AI Technical Support Specialist (Jun–Jul 2026) | The AI agent handles 85%+ of conversations; specialists take complex cases and can grow into managing the agent |
| Pattern | Marketplace Account Executive (open) | Claude daily for account research, plans and CRM upkeep; a book of about 100 accounts |
| Skai | Agentic Marketing Operations Manager (Apr 2026) | Agents run data enrichment, lead routing and syncs between tools |
| Gelato | AI Marketing Specialist (Jun 2026) | Builds agents for search-term mining, budget pacing, anomaly checks and reporting |
| VTEX | Product Support Engineer, Expert Hub (Aug 2026) | Builds agents and MCP servers to add second-line support capacity |
Compare: Pacvue account managers build client reports in Excel, Gorgias plans to scale billing support to 4k tickets a month with people, and Threecolts reconciles payment systems by hand.
What trigger employees say about AI
A skeptical buyer will check. Across ~76 staff reviews, only 4 mention AI; the before/after numbers come from leaders and case studies, not employees.
| Trigger | Glassdoor | What staff say | Use as proof |
|---|---|---|---|
| Omnisend | 4.6 · 91% recommend | Nothing about AI yet | Safe |
| Gelato | 4.2 · 77% | One long-tenured employee praises AI and automation (Apr 2026) | Safe |
| Hostinger | 3.9 · 72% | AI tools praised; some customer success staff fear cuts despite the company's "no one lost their job to AI" | Use, expect questions |
| VTEX | 3.8 · 69% | Repeated layoffs and understaffed teams in 2025–2026; CEO says support went from 120 to 45 people with NPS up 11 points | Use the numbers, not the culture |
| Pattern | 4.3 · 86% | Nothing about AI | Job posts only |
| Yotpo | 3.4 · 46% | Remaining staff working "twice as hard" after cuts (Jan 2026) | Avoid as lead example |
| Skai | 3.4 · 59% | Restructuring every couple of months, older than the AI story | Avoid as lead example |
Takeaway for later: target employees already describe layoffs and overload, and at Yotpo, Skai and VTEX "AI means layoffs" is what a skeptic will find. Relief for overloaded teams is a safer story than headcount cuts.
Samples and limits
- Customer reviews: Helium 10, 50 most recent 1–2★ Trustpilot reviews (Mar 2025 to Jul 2026). Threecolts, 23 reviews of 1–3★ for the parent and its InventoryLab, Tactical Arbitrage and FeedbackWhiz brands. Teikametrics, all 17 Trustpilot reviews of 1–3★ plus G2. Gorgias, 40 Shopify App Store reviews of 1–3★ (2025–2026). Jungle Scout, 12; Carbon6, 7; Quartile, too few to count reliably.
- Employee reviews: Glassdoor shows about 3 reviews per page without login, so samples were built from its keyword pages (layoffs, workload, management) and lean negative. Treat counts as direction, not as rates. Carbon6's own reviews were unreadable; that column is its owner, SPS Commerce. RepVue quota figures are self-reported by sales reps.
- Job posts: most show no date; "open" means the page still took applications on 12 Oct 2026.
- Proof quality: most "who fixed it" examples are vendor case studies (Intercom Fin, Glean, Guru, Notion, Tabs) that name the company and give a number; several are undated. Lightspeed (earnings call), Intercom and Hostinger (own sources) are the strongest.
- Not readable: Reddit and most of Blind.
Step 3 · Task
Turn the problems AI can fix into a few offers, decide which target hears which offer first, and write the ads and 1:1 messages that bring in leads.
Inputs
- AI-fixable problems from step 2, with counts and evidence
- Proof from triggers and case studies: Hostinger, Omnisend, AMZ Advisers, Cortex and others
- Each target's job posts and public reviews, for 1:1 hooks
- Rules: no prices, capacity rather than headcount cuts, every number linked to its source
Expected result
- A small set of offers, each with what we build, what we measure in the pilot and proof
- One sales path for every offer: audit, then pilot, then program
- Which offer leads for which target, with the hook
- Ad creatives per buyer role and a 1:1 message per target
- Messaging rules for public ads and for direct messages
What we found
Three offers cover 7 of the 9.5 problems AI can fix. Each is sold the same way: an audit, then a pilot on one workflow, then a program, with the number we will move agreed before the pilot starts. Pricing is for Oberon to set.
How every offer is sold
The same three steps for each offer. Each step ends in a decision the client makes on its own numbers.
- Audit2 weeks
We read the real tickets, reports or billing data. The client gets its own value figure and a ranked roadmap.
- Pilot30 to 45 days
One workflow live on real work, with the number we will move agreed before it starts.
- Program90 days and on
Every queue or team, an owner and monthly metrics; then the next function.
Who gets which offer first
| Target | Lead with | The hook | Then |
|---|---|---|---|
| Pacvue (Helium 10's owner) | 1 · Copilot | ~10 open account manager roles building reports in Excel | 2 for Helium 10 |
| Helium 10 | 2 · Support | slow support or bot-only replies come up 20 times in 50 recent low reviews | 3 |
| Quartile | 1 · Copilot | managers carry 15+ accounts, QBRs in Excel | – |
| Teikametrics | 1 · Copilot | CSMs prepare recurring business reviews; a 10-year client lost to AM silence | 2 |
| Jungle Scout | 1 · Copilot | CSMs with 40–60 accounts update health scores and CRM by hand | – (support is rated well) |
| Threecolts | 3 · Billing | reconciles 5+ payment gateways to Sage every month | 2 |
| Gorgias | 3 · Billing | billing tickets planned to grow from 1k to 4k a month | 1 (onboarding load) |
| Carbon6 | Hold | Confirm who owns which brands after the June 2026 sale | 2 |
Gorgias is not offered support automation first: it sells AI support itself and is staffing its own AI-native CX team.
Offer 1
Account manager copilot
Report and QBR prep in minutes instead of hours, and an account history that stays when people leave.
Closes: manual client reports · account managers who change or go quiet · uneven onboarding · knowledge loss
What we build
- Report and QBR drafts: an agent pulls ads and marketplace data and drafts the weekly report and the QBR deck; the account manager reviews and sends.
- Account memory: every call, change, promise and open item in one place, so a new manager picks up without the client repeating themselves.
- Onboarding checklist: connects accounts, sets up first campaigns from a template, flags what is missing.
What we measure in the pilot
- Minutes per client report, before and after
- Accounts per manager
- Days from signature to first live campaign
- Client emails waiting more than 48 hours
Proof
- AMZ Advisers, an Amazon agency with 85+ accounts: reports from 2–3 hours to 30–40 minutes, 120–150 hours a month saved
- Storable: account audit from 3–4 hours to about 10 minutes · Abnormal: QBR prep to 5–10 minutes · Performio: onboarding from 20 weeks to 3.5
Creatives
Ad · CEO or PE operating partner
Your account managers are spreadsheet operators.
An Amazon agency with 85+ accounts lets AI draft each monthly client report and has people review it: 30–40 minutes instead of 2–3 hours, and 120–150 hours a month back for clients.
See it on one of your reports
Ad · Head of client services
When an account manager leaves, the account history shouldn't.
Every call, change and promise in one place, so the next manager picks up mid-conversation. Storable cut account audits from 3–4 hours to 10 minutes.
Book a 20-minute walkthrough
Offer 2
Support that resolves
An agent that closes tickets instead of deflecting them, coached by your best people.
Closes: slow support and bots that loop · overloaded teams · unanswered reviews (add-on)
What we build
- An agent on the help desk you already use (Zendesk, Intercom, Gorgias), connected to account and billing data so it can act: check a plan, process a cancellation under your policy, fix a sync.
- Tested on the last three months of real tickets before launch; it goes live only on ticket types where it beats the current bot.
- A weekly coaching loop where support leads review misses and update the agent. Add-on: drafted replies to public reviews, approved by a person.
What we measure in the pilot
- Share of tickets resolved without a person
- Satisfaction for AI-resolved vs human-resolved tickets
- First response time
- Cancellation requests waiting more than 24 hours
Proof
- Hostinger: 91% of ~1.5M monthly conversations close without a person; specialists coach the AI on the hard ones
- Omnisend: ~40% of tickets, satisfaction level with humans after two people coached the agent · Intercom: 81%, about 100 hires avoided
Creatives
Ad · CEO or PE operating partner
Your customers don't hate AI support. They hate AI that can't fix anything.
Omnisend's agent started at a third of human satisfaction. Two people coached it, and on ~40% of tickets it now matches the team.
See which of your tickets an agent could close
Ad · Head of support
Your team should coach the AI, not clean up after it.
At Hostinger, 91% of conversations close without a person, and specialists spend their time training the agent on the hard cases.
Get a ticket audit
1:1 · Helium 10
Slow replies or bot-only answers come up 20 times in Helium 10's 50 most recent 1–2★ Trustpilot reviews (reviews). Hostinger now closes 91% of conversations without a person, with specialists coaching the AI on the rest (source). Open to a 20-minute look at which tickets to start with?
Offer 3
Billing and collections desk
Billing tickets, reconciliation and payment reminders handled without adding finance or support headcount.
Closes: billing ops and collections by hand. It applies your refund policy consistently; it does not change it.
What we build
- A billing-ticket agent connected to Stripe or Chargebee: answers invoice questions, applies the existing refund policy, sends exceptions to a person with the full context.
- A reconciliation assistant that matches payment gateways to the ledger every day and lists only the mismatches.
- Collections follow-up: reminders drafted by how overdue the invoice is, escalated by your rules.
What we measure in the pilot
- Billing tickets resolved without a person
- Days to collect
- Hours spent on month-end reconciliation
- Chargebacks
Proof
- Cortex, a 2-person finance team: days to collect from ~45 to 25, month-end close from ~7 days to 3, overdue invoices halved
- Omnisend: invoice processing up to 85% automated · Gamma: 75% of refund and failed-payment requests resolved end to end
Creatives
Ad · CEO or PE operating partner
Billing tickets grow with revenue. Your billing team doesn't have to.
A two-person finance team cut days to collect from 45 to 25 and month-end close from 7 days to 3 with AI invoicing and follow-ups.
Estimate your billing hours
Ad · Controller or head of finance
Stop reconciling payment gateways by hand.
An agent matches every gateway to your ledger daily and shows you only the mismatches. Omnisend's finance team automated up to 85% of invoice processing.
See it on one month of data
Messaging rules
- Sell relief and capacity, not cuts. Target employees already describe layoffs and overload, and at Yotpo, Skai and VTEX a skeptic will read AI as layoffs. "Your team stops doing X by hand" lands; "fewer people" does not.
- Public ads: never name a target or quote its reviews. Use the problem and a trigger's number.
- 1:1 messages: public review counts and job posts are fair to cite, with the link. Employee reviews shape which offer we lead with, but are never quoted to the company.
- Every number links to its source. Lead examples: Hostinger, Omnisend, AMZ Advisers, Cortex. Use Yotpo and Skai only with their caveats.
- Held back for now: a sales research and follow-up offer (evidence at only 3–4 targets, crowded market) and review replies as a standalone product (better as an add-on to Offer 2).
Step 4 · Task
Turn people who respond to our ads and messages into leads we can send a report to: a work email, a company, and a reason to believe the company fits.
Inputs
- Offers, ad creatives and 1:1 messages from step 3
- The target list and the target bar from step 1
- Channels: LinkedIn ads with a native lead form, other ads and email with a landing page, 1:1 messages
Expected result
- A lead record: name, work email, company, role, and the offer and creative that brought them
- A fit check for every lead against the target list and the target bar
- Every qualified lead gets their report within 1–2 business days
- Numbers per creative: responses, qualified leads, reports opened, call clicks
How it works
Landing pages: one per offer, not one per creative
Three pages cover every creative. The link in each ad adds ?c= with the creative's id, and the page swaps its headline and first line to match the ad, so the visitor sees the promise they clicked on. LinkedIn traffic uses the native lead form and skips the landing page. The form asks for a work email, the company website and a role; it rejects free email addresses and keeps the offer, creative and UTM tags for attribution.
| Offer | Creative for | Headline the visitor sees | Link |
|---|---|---|---|
| Support that resolves | CEO or PE operating partner | Your customers don't hate AI support. They hate AI that can't fix anything. | landing-support.html?c=support-ceo |
| Support that resolves | Head of support | Your team should coach the AI, not clean up after it. | landing-support.html?c=support-head |
| Account manager copilot | CEO or PE operating partner | Your account managers are spreadsheet operators. | landing-copilot.html?c=copilot-ceo |
| Account manager copilot | Head of client services | When an account manager leaves, the account history shouldn't. | landing-copilot.html?c=copilot-cs |
| Billing and collections desk | CEO or PE operating partner | Billing tickets grow with revenue. Your billing team doesn't have to. | landing-billing.html?c=billing-ceo |
| Billing and collections desk | Controller or head of finance | Stop reconciling payment gateways by hand. | landing-billing.html?c=billing-finance |
For software companies that serve Amazon and ecommerce sellers
An agent that closes tickets instead of deflecting them, coached by your best people.
We look at your company's public reviews and job posts and show what an agent could take off your support team, what that is worth a year, and who has already done it.
In your free report
- Where it hurtsWhat your customers and staff say in public, and which of it AI can fix
- What it is worthThe capacity AI could free on your support team, with a calculator for your own numbers
- Who has done itSoftware companies that already moved the same work to AI, with sources
What you get
One page on your own company, not a generic whitepaper
Built from your public reviews, job posts and peers. Numbers you can change to match your team, and every fact linked to its source.
Who has already done it
Software companies that moved the same work to AI
of ~1.5M monthly conversations closed without a person; specialists coach the AI
Hostingerof its own support resolved by AI; about 100 support hires avoided
Intercomof tickets at the same satisfaction as people, after two staff coached the agent
OmnisendMarket examples, not our clients.
For software companies that serve Amazon and ecommerce sellers
Report and QBR prep in minutes instead of hours, and an account history that stays when people leave.
We look at your company's public reviews and job posts and show how much report and account work AI could take off your account managers, what that is worth a year, and who has already done it.
In your free report
- Where it hurtsWhat your customers and staff say in public, and which of it AI can fix
- What it is worthThe capacity AI could free on your account managers, with a calculator for your own numbers
- Who has done itSoftware companies that already moved the same work to AI, with sources
What you get
One page on your own company, not a generic whitepaper
Built from your public reviews, job posts and peers. Numbers you can change to match your team, and every fact linked to its source.
Who has already done it
Software companies that moved the same work to AI
per monthly client report instead of 2–3 hours, across 85+ Amazon accounts
AMZ Advisersper account audit instead of 3–4 hours
Storableof QBR prep
AbnormalMarket examples, not our clients.
For software companies that serve Amazon and ecommerce sellers
Billing tickets, reconciliation and payment reminders handled without adding finance or support headcount.
We look at your company's public reviews and job posts and show how much billing and collections work AI could take off your team, what that is worth a year, and who has already done it.
In your free report
- Where it hurtsWhat your customers and staff say in public, and which of it AI can fix
- What it is worthThe capacity AI could free on your billing team, with a calculator for your own numbers
- Who has done itSoftware companies that already moved the same work to AI, with sources
What you get
One page on your own company, not a generic whitepaper
Built from your public reviews, job posts and peers. Numbers you can change to match your team, and every fact linked to its source.
Who has already done it
Software companies that moved the same work to AI
days to collect for a two-person finance team; month-end close from ~7 days to 3
Cortexof invoice processing automated
Omnisendof refund and failed-payment requests resolved end to end
GammaMarket examples, not our clients.
Who gets a report
- The company is on our target list, or meets the target bar: PE or VC money, roughly $25–150M revenue, operations still run by hand
- The person can buy or sponsor the work: CEO, PE operating partner, VP Customer Experience, head of account management or client services, CFO or controller
- A work email at the company's domain; the landing page asks again when someone enters a free email address
- Competitors, agencies and job seekers get a polite reply and no report
What we measure
- Per creative: click-through rate and form completion rate
- Share of leads that pass the fit check
- Time from the form to the report being sent
- Per report: opened, calculator used, call button clicked (each lead gets their own link)
Step 5 · Task
Give each interested lead one page about their own company: what we found from public data, how much team time AI could free, and a reason to book a call.
Inputs
- The lead's company and its evidence from steps 1–2: maturity scores against peers, review counts, job posts, ratings
- Market examples from steps 2–3, labelled as not our clients
- Public-data assumptions for a low, middle and high case: team sizes, loaded costs, AI share, time saved
Expected result
- A page that shows in 30 seconds where the gap is, what it costs and the next step
- A calculator the lead can set to their own numbers, with every figure following
- Every external fact linked to its source; problems AI does not fix stay visible
- One self-contained file that opens on a phone and prints to PDF, and a booked call
The result
What an interested lead receives. Our findings from public data, a calculator for their own numbers, and a button to book the call. This is the same file we send.
Helium 10: the business case for AI in customer operations
Before our call · built from public information
Your estimate · capacity freed a year
$853,000 a year
Public data puts it at $0.24M–$2.3M. Change the inputs to match your team.
Helium 10 ships AI to sellers, but the customer operation behind it still runs on people and spreadsheets. On public evidence alone, AI in support and in Pacvue's account-manager reporting is worth about $0.24M–$2.3M a year in freed capacity, with a middle case near $0.85M.
Where the gap is
Support and account-manager reporting, which still run on people and spreadsheets.
Why now
Helium 10 is hiring a VP Customer Experience whose brief is to decide where AI and where people serve customers.
Next step
A 30-minute call to check these numbers against your team's real figures.
Book a 30-minute call01What we already know
Public evidence shows the gap sits in support and in managed-services reporting, not in the product.
1of 5
AI in operations: the lowest score of the five companies compared. AI in the product scores 2.5.
2.1of 5
Trustpilot rating from 712 reviews, with no reply to a negative review since April 2025.
~10roles
Open Pacvue account-manager roles that build weekly and monthly client reports, mainly in Excel and Sheets.
2.9of 5
Glassdoor rating, 47% would recommend; the same work spread over fewer people is a recurring theme.
What the 50 most recent 1–2★ reviews name
Where Helium 10 stands against peers
02What AI could free up
Even the low case is worth more than $200k a year in team capacity.
The three cases use only public signals and peer benchmarks. Start from one, then change any input to match your team; every figure on this page follows.
Capacity freed a year, by case
All cases side by side
| Line | Low | Middle | High | Yours |
|---|---|---|---|---|
| Support | ||||
| Support agents | 10 | 20 | 30 | 20 |
| Loaded cost per agent, $ a year | 35,000 | 50,000 | 70,000 | 50,000 |
| Share of tickets AI resolves after 6 months | 40% | 50% | 60% | 50% |
| Capacity freed, full-time agents | 4 | 10 | 18 | 10 |
| Support value, $ a year | 140,000 | 500,000 | 1,260,000 | 500,000 |
| Account-manager reporting | ||||
| Account managers building reports | 20 | 35 | 50 | 35 |
| Hours a week on reports and QBR prep | 5 | 7 | 10 | 7 |
| Time AI drafting removes | 70% | 75% | 80% | 75% |
| Working weeks a year | 48 | 48 | 48 | 48 |
| Hours freed a year | 3,360 | 8,820 | 19,200 | 8,820 |
| Loaded cost per hour, $ | 30 | 40 | 55 | 40 |
| Reporting value, $ a year | 100,800 | 352,800 | 1,056,000 | 352,800 |
| Total capacity, $ a year | 240,800 | 852,800 | 2,316,000 | 852,800 |
Where the assumptions come from
- AI share of tickets: peers report 40% (Omnisend) to 80–91% (Lightspeed, Hostinger); we stay at the low end because customers already dislike the current bot.
- Tickets stand in for time: we count the share of tickets AI resolves as the share of agent time it frees. Routine tickets are quicker than bugs, so with your real tickets each type is weighted by its handle time.
- Time saved on reports: an Amazon agency cut each monthly client report from 2–3 hours to 30–40 minutes, about 75–80% (AMZ Advisers).
- Team sizes: Helium 10 has about 190 staff (PitchBook); the support and account-manager headcounts are our assumptions.
- Working year: 48 working weeks of 40 hours.
- Run-rate, not year one: the figures apply once AI reaches its six-month share. Year one is lower while it ramps up.
- Capacity, not cuts: the value can go into growth, such as the planned managed-services department, instead of new hires.
03What others achieved
Companies that moved the same work to AI report results in the range we use above.
Share of support resolved by AI
Time per task, before and with AI
Market examples (not our clients)
| Company | What they did | Result | Source |
|---|---|---|---|
| Hostinger | AI agent on its own support; specialists coach it | 91% of ~1.5M monthly conversations closed without a person | Sep 2026 |
| Intercom | Its own AI agent on its own support | 81% resolved; about 100 support hires avoided | Mar 2026 |
| Lightspeed | AI on support tickets | More than 80% resolved, per the CFO | May 2026 |
| Omnisend | Agent trained by two staff | ~40% of tickets at the same satisfaction as people | Sep 2026 |
| AMZ Advisers | AI drafts monthly client reports for 85+ Amazon accounts | 2–3 hours down to 30–40 minutes per report; 120–150 hours a month saved | Apr 2026 |
| Storable | AI agent audits CSM accounts | 3–4 hours down to about 10 minutes per audit | Glean case |
Next step
Let's check these numbers against yours.
Thirty minutes with Oberon.ai. We go through what we found and replace our assumptions with your team's real figures.
Book a 30-minute callOn the call
- Your real numbers in the calculator, in place of our public-data assumptions
- Which queue or workflow would give your team the most time back first
- What a first step would look like, and what it needs from your side
04
Sources
- Helium 10 Trustpilot profile · https://www.trustpilot.com/review/www.helium10.com
- VP Customer Experience posting · https://job-boards.greenhouse.io/pacvue/jobs/6201012004
- Pacvue Retail Media Data Account Manager posting · https://job-boards.greenhouse.io/pacvue/jobs/5451313004
- Helium 10 on Glassdoor · https://www.glassdoor.com/Reviews/Helium-10-Reviews-E2117134.htm
- Helium 10 on PitchBook · https://pitchbook.com/profiles/company/435272-32
- Hostinger: support cases handled by AI · https://webhosting.today/2026/09/09/most-of-hostingers-hardest-support-cases-no-longer-transfer-to-a-human-the-human-coaches-the-ai-instead/
- Intercom: automating its own customer service · https://www.intercom.com/blog/automate-customer-service-while-improving-customer-experience/
- Lightspeed Q4 earnings call highlights · https://www.nasdaq.com/articles/lightspeed-commerce-q4-earnings-call-highlights
- Omnisend on its support agent (podcast) · https://podscan.fm/e/ep_y2lkxde49w6yjedo
- AMZ Advisers: AI across 85 Amazon accounts · https://mikebegg.me/blog/how-we-use-ai-to-manage-85-amazon-accounts
- Storable customer story (Glean) · https://www.glean.com/resources/customer-stories/storable