Talomnia
RU

Research / market-research-talomnia-workforce

Talomnia Workforce — Market Research

marketpricingunit-economicsicpself-use

Everything this study concludes is built on the tables below, and the counts here are computed from those tables — nothing is summarised away and every row remains readable.

data tables in this study
2
rows across them
18
sources cited
46

Category landscape, competitors and the category's documented failure modes, price benchmarks across thirteen adjacent categories, the first self-use unit-economics measurement from the Workflow Ledger, the ICP and the first sellable service. Built by Talomnia for Talomnia — pre-commercial self-use validation: produced by the Talomnia agentic workforce for Talomnia itself; there are no commercial cases yet.

Evidence record

What is verified
2 tables, 18 data rows, 46 sources
Revision
Version 0.2.0 dated 2026-08-19
Validation status
Published research from a self-use case; publication is not independent validation.

Executive summary

Talomnia Workforce sells completed professional work — research, software development, technical and product analysis — with the economics fixed before execution and an evidence trail after it. This research answers three questions: does the category exist, who solves the same tasks today and at what price, and what to sell first. It was produced by the Talomnia agentic workforce for Talomnia itself (Built by Talomnia for Talomnia — pre-commercial self-use validation) and is itself a demonstration of the sellable service.

  • The "completed work instead of a tool" category is already being validated commercially — in narrow verticals; across professional domains the slot is unclaimed.
  • Buyers pay for accountability and verifiability, not labor content: human rates for the same work span three orders of magnitude.
  • Sell first: fixed-scope research and technical-analysis engagements with acceptance criteria and a Budget Limit — and the AI-first Transformation Consultation.

Back to contents

ICP — who buys first

The primary ICP (our hypothesis — derived, not surveyed, and falsifiable by the first commercial engagements): a founder, C-level or head of operations at a 10–200-person technology-adjacent company that already uses AI tools and believes in agentic execution but is not scaling agents internally; whose professional tasks arrive faster than it can hire; and which knows the category's documented failure modes, so evidence-backed delivery is a feature it can evaluate. The buying trigger is a concrete deferred task, not an "AI transformation" ambition.

The secondary ICP: mid-market companies whose leadership needs a paid, bounded entry point into AI-first transformation — a concrete assessment instead of a platform commitment; boutique assessments at $10–35k are an established purchase in this segment. The explicit non-ICP for the validation stage: enterprises with long vendor-onboarding cycles and price-only buyers comparing against a tool subscription (a category error).

Back to contents

Customer pains — what the buyer is escaping

  • Hiring is slow and expensive while tasks are already queued: work arrives faster than hiring; "capability without hiring" is a growing purchase (48% of CEOs plan more freelance use; 81% shift to skills-based hiring).
  • Outsourcing disappoints yet persists: only 25% of executives report lower cost or higher quality from AI-powered outsourcing; 70% repatriated some work within five years — and still 80% keep or grow the spend. Demand persists; satisfaction does not.
  • Tools transfer the work to the buyer: metered agent access leaves prompting, orchestration and verification with the client; acceptance stays their problem.
  • Burned by fabrication: after the category's public failures, buyers actively verify claims; unverifiable marketing is now a negative signal.
  • Ambiguity of "what counts as done": attribution disputes and unclear success criteria are the top named risk of buying outcomes; pre-agreed acceptance criteria with an evidence trail are what closes it.

Back to contents

Market landscape

The category boundary drives every comparison: Talomnia sells an accepted outcome, not model access or a seat licence. Comparing Talomnia's price to a tool subscription is a category error in both directions: the tool is cheaper because the buyer still does the professional work; an outcome service compares to the cost of the same accepted result from a human performer.

The "AI-executed completed work" category already exists and grows in narrow verticals: outcome pricing is now standard in customer support, an AI-native law firm sells finished contract reviews at a fixed fee, hybrid operations-as-a-service more than doubles revenue year over year, and institutional capital explicitly bets on buying service firms to re-base them on AI execution. Nobody was found selling Talomnia's full bundle — Budget Limit + acceptance criteria + evidence trail + rework guarantee — across domains, in a bounded search (~15 queries); that is absence of evidence, not proof of absence.

Back to contents

Direct competitors

  • Sierra — the category's proof point: $100M ARR within ~21 months on customer-service agents with explicitly outcome-based pricing (in most cases the customer does not pay when the agent hands work to a human) [primary/secondary].
  • Crosby — the closest structural analogue: an AI-native law firm selling completed contract reviews at a fixed per-document fee (~$400 typical) with sub-hour turnaround; humans + AI inside one accountable entity [secondary].
  • Invisible Technologies ($134M revenue in 2024, doubling YoY) and Superside ($44.9M revenue in 2024 on subscription finished creative work; ARR estimates conflict across sources) — proof that productized completed work scales [secondary/primary].
  • The contrast: Cognition/Devin — the "AI engineer" — is sold as a metered tool (Agent Compute Units), not as accepted deliverables; the quality criticism lands exactly where the buyer owns acceptance. The gap between metered access and an accountable accepted outcome is where Talomnia positions.

Back to contents

Adjacent alternatives — what the client buys today

  • Consulting: ~a quarter of McKinsey's global fees are reportedly outcome-based; only about half of clients believe consultants create value beyond what was paid for [secondary].
  • Outsourcing/BPO: demand persists amid chronic dissatisfaction (Deloitte 2024 [primary]) — see Pains.
  • Freelance marketplaces: Upwork and Fiverr both show fewer buyers at higher spend [primary] — the low end of gig work is absorbed by AI tools; what remains bought is bigger and trust-priced.
  • On-demand research: Wonder — the pre-AI "completed research at a fixed price" service — dissolved as of September 2025 [secondary]; direct evidence that a bare research deliverable without accountability does not survive the deep-research-tool era. Expert networks (GLG, AlphaSights) persist at premium prices.
  • Fractional executives: a growing "capability without hiring" adjacency, sized $8.6–9.4B for 2025 — low-grade sources that disagree by ~40% [secondary, low confidence].

Back to contents

Pricing benchmarks in adjacent categories

The client compares Talomnia not with a tool subscription but with the alternative cost of the same accepted result. All values were accessed 2026-08-19; source grades as in Methodology; full wording and caveats live in the canonical research document (linked from the Workflow).

CategoryObserved rangeSource grade
Freelance marketplace, hourly (Upwork)developers median $20/hr (typical $10–100); market research analysts $25–70/hr; business consultants median $55/hrprimary via snippet — re-verify before reuse
Curated talent network (Toptal)no published rate card; secondary sources converge on $60–150+/hrsecondary, directional only
Offshore staff augmentation2026 sweep — LatAm senior $60–75/hr; Asia senior $31–41/hr; CEE senior $64–76/hrprimary (industry rate survey)
Software dev agenciesmost $24–49/hr; span up to $150+/hrprimary via snippet
Management consultingMBB partners $500–1,000+/hr; Big Four UK £800–1,500/day junior to £3,500–6,000/day partnersecondary — no firm publishes rates
Custom market research projectstypical custom study $25,000–65,000; qualitative ~$8k–22.5k per 10–15 interviewsprimary for vendors' own pricing, secondary as market norm
Expert networks~$600–2,500 per hour or session; no published rate cardssecondary — quote-only
Analyst reportsForrester single reports $2,995 (some $1,495); subscriptions $25k–500k+/yrprimary (forrester.com) / secondary
AI agent products (tool-metered)Devin $20/mo entry + ~$2.25/ACU (≈ $9 per metered agent-hour); ChatGPT Pro up to $200/mo (top tier)secondary / widely published
AI "employee" products (subscription)11x Alice $36k–45k/yr (the vendor page is internally inconsistent); Lindy $29.99–199.99/user/moprimary (pricing pages)
AI outcome pricing (support vertical)$0.99–2.00 per resolved conversation; Sierra contracts from ~$150k/yrsecondary
AI readiness / transformation assessmentsboutique fixed-fee $10k–35k; Big-Four-tier $50k–150kprimary (boutique rate cards) / secondary

Structurally: human professional work spans three orders of magnitude — buyers pay for vetting, verification and someone to hold responsible; outcome pricing arrives from two directions (AI-native vendors and consulting); metered agent access gets cheaper while outcome-accountable delivery holds premium — value migrates to verified completion.

Back to contents

Unit economics — the first self-use measurement

The numbers below are measured from the Workflow Ledger of this very launch epic — Talomnia's agents building Talomnia. They are the first real datapoints of the cost structure of agent-executed work in this system, and NOT commercial unit economics: no client, no price, no external acceptance; no against-human cost coefficient is derived from them.

MetricMeasured value
Wall time, total48 min 18 s
Active execution time40 min 59 s
Tokens (in / out)570,500 / 85,500
Model cost (estimated)≈ $13.10
Rework rate2 of 22 entries (9.1%)
First creation vs reuse≈ $1.67 to create an artifact vs ≈ $0.18 per reuse

The cost profile is dominated by first creation (≈ nine times the cost of each later reuse — the mechanism the Talomnia thesis rests on, observed for the first time in our own ledger); rework is measurable and an honest price must carry a reserve for it; model compute is a minor share of what a client would pay for equivalent work, so price anchors to the alternative cost of the result, not to inference.

Back to contents

Buying behaviour

  • The C-suite already buys work-without-hiring: about half of CEOs plan to increase freelance use; 51% say the business would be hard to run without freelancers; top performers adopt hardest [primary via coverage].
  • Agent adoption sits at the experimentation-to-scaling transition: 88% of organizations use AI in at least one function, 62% experiment with agents, only 23% scale them [primary] — most want agentic execution and cannot yet operate it themselves; that is exactly the buyer of delivered outcomes.
  • The marketplace pattern concentrates: fewer buyers, larger engagements — small transactional gigs migrate to AI tools; what remains purchased is bigger, higher-stakes work [primary].

Back to contents

Market risks — the category's documented failure modes

  • Fabricated traction / AI-washing: the Builder.ai collapse (claimed $220M revenue against ~$55M audited; the "AI" turned out to be hundreds of developers) and the 11x exposure (logos of non-customers, inflated ARR) taught buyers to verify claims [secondary/primary].
  • Quality variance and churn: AI SDR products report 50–70% annual churn [secondary; the figure's upstream attribution is unconfirmed] — selling outcomes without a verification layer produces churn.
  • "Defining done": the hardest part of outcome pricing is a contractually unambiguous, technically measurable definition of done [secondary]; exactly what acceptance criteria + an evidence trail address.
  • Commoditization of the bare deliverable: deep-research tools collapse the price of an unaccountable report toward a subscription; Wonder's death is the market's verdict. What stays scarce is accountable, verifiable, accepted work.

Back to contents

The most promising first wedge

The "accountable, sourced, accepted research deliverable" slot is open right now: the niche's incumbent dissolved, and the surviving alternatives are either premium expert-hours ($500–2,000/hr) or unaccountable deep-research tool output. The predecessor's death is priced in: the offer must sell accountability and acceptance, not the bare report.

Back to contents

Recommendation: the first sellable service

  • First offer: fixed-scope research and technical-analysis engagements (market research, technical due diligence, architecture and product analysis) with pre-agreed acceptance criteria, a Budget Limit and an evidence trail — the capability this very document demonstrates, and a deliverable verifiable by reading.
  • Second offer in the same window: the AI-first Transformation Consultation — a bounded paid assessment with the same evidence discipline; the $10k–35k boutique-assessment benchmark shows an established purchase pattern (Talomnia's price is an operator decision; the recommendation was delivered separately).
  • Deliberately not first: end-to-end software delivery (highest stakes and the hardest acceptance definition — start after research-class trust is established) and anything sold as an "AI employee" subscription (the category's churn pattern).

Back to contents

Goal

Ground two launch decisions: the recommendation on future Workforce rates and unit economics (the rates themselves are not published — an operator decision) and the recommendation for pricing the AI-first Transformation Consultation; plus the § 5.6 specification mandate — identify the ICP, pains, competitors, prices, alternatives and the first sellable service for the next 30 days.

Back to contents

Methodology

Desk research on 2026-08-19: two independent passes over public sources (competitor/category landscape and price benchmarks); every market figure carries its source, access date and grade — [primary] seen on the owner's own page, [secondary] press/analyst/aggregators, [snippet] visible only in search results; anything unverifiable is recorded as a gap, not guessed. The primary self-use measurement comes from this launch's own Workflow Ledger (time, tokens, cost, rework share). Before publication every published source was actually fetched and recorded in a verification manifest; sources that did not resolve are named.

Back to contents

Results

The category "we sell completed work, not a tool" is real and is being validated commercially by others in narrow verticals; the slot "verifiable accountable work across professional domains" is unclaimed. Trust is the battleground: the market has just learned the price of fabricated claims. Pricing is migrating from model access to the accepted outcome. The first own unit-economics datapoints are measured: creating a knowledge artifact cost ≈ nine times its later reuse, rework share 9.1% (a small internal sample, not a marketable coefficient).

Back to contents

Limitations

No commercial cases — all unit economics is a pre-commercial self-use measurement; market figures are third-party and exactly as good as their sources (grades are inline per figure); ledger costs are token-based estimates at list prices, not invoices; the sample is 22 entries over 4 infrastructure tasks; the category moves fast — the 2026-08 pricing sweep must be re-run before any rate decision.

Back to contents

Conclusions for Talomnia

Talomnia enters an existing category whose cross-domain generalization is unclaimed; the differentiation to lead with is the Evidence + acceptance-criteria + Budget Limit bundle, which answers the category's documented failure modes one-for-one — and works only while the honesty constraint holds: no claimed traction, targets marked as targets. Price should follow the outcome, not the hour of compute. First 30 days: sell fixed-scope research/technical-analysis engagements and the AI-first Transformation Consultation to the primary ICP, with this document and the launch Workflows as the demonstration artifacts.

Back to contents

Sources

Back to contents

How this was performed

/en/workflows/market-research-talomnia-workforce

Back to contents

Versions

v0.2.0 — 2026-08-19