Revenue Model and Pricing
Revenue model, pricing tiers, unit economics, and financial projections for the Sókrates outsourced AI department.
5. Business Model
5.1 Revenue Model
Sókrates generates revenue through a single stream: a monthly retainer per customer company. AI consumption (Claude Teams seat costs) is passed through at cost with full transparency. There is no token markup, no usage metering, no hidden fees. The retainer is the revenue. The seats are a passthrough. The on-premises hardware (CWWK N305; see CWWK 4-LAN N305 (Sokrates Box)) is a one-time onboarding cost rolled into the first month — the customer owns the box from day one.
The retainer scales with two dimensions: company size (more employees means more workflows, more MCP connectors, more surface area for the Sókrates Agent to operate on, and more Claude Teams seats) and committed weekly hours of founder/team time (larger or more complex organizations require more human attention for validation, edge cases, strategic decisions, and governance that the Sókrates Agent can’t handle alone).
Critically, the retainer does not scale with plugin count. The Sókrates Agent finds whatever it finds and builds whatever it builds. Plugins are unlimited within the engagement scope. The customer is not buying automations — they are buying an AI department. You don’t pay your department more because they had a productive month. Tiering by output count would create a perverse incentive where the Sókrates Agent’s best feature — proactive discovery — becomes a billing trigger the customer learns to dread. Aligned incentives mean the agent’s productivity is the customer’s gain, not their invoice.
Pricing logic anchored to alternative cost:
The customer’s alternative to Sókrates is some combination of: hiring an internal AI specialist, engaging a consulting firm for periodic projects, or continuing with ungoverned ChatGPT subscriptions (see §3.3 for the cost analysis of each alternative). Against this backdrop:
Sókrates retainer — ISK 400–800K/month depending on company size and committed time
Every customer gets the full stack from day one. There is no feature-gated tier. The Sókrates Agent runs in proactive discovery mode. The box is deployed with Eidos and customer-selected MCP connectors. The base bundle is live. Plugins are built as the Sókrates Agent identifies opportunities and are maintained for the life of the engagement. Governance monitoring is active. The only variable is the intensity of the engagement: how many hours per week of founder/team time are committed to this customer’s operations.
One-time onboarding fee: ISK 30K per seat purchased. Covers Eidos seeding from customer documentation, MCP connector configuration, trust boundary co-design with CEO/CFO, NixOS image deployment, Sókrates Agent initial calibration, base bundle configuration, employee onboarding program, and hardware (CWWK N305, included). The fee scales naturally with company size because larger companies have more documentation, more systems to connect, and more employees to onboard. For EDIH-IS subsidized trials, the onboarding fee is covered by the subsidy.
Indicative pricing by company size:
25–35 employees (simpler operations, fewer systems): ISK 400K/month retainer + ISK 35–55K/month Claude Teams seats (12–18 seats). Monthly ongoing: ISK 435–455K (approximately €2,900–3,050/month). Onboarding: ISK 360–540K. Committed time: 3–5 hours/week.
35–55 employees (moderate complexity, multiple departments): ISK 600K/month retainer + ISK 50–85K/month seats (18–28 seats). Monthly ongoing: ISK 650–685K (approximately €4,350–4,550/month). Onboarding: ISK 540–840K. Committed time: 5–8 hours/week.
55–75 employees (high complexity, cross-functional workflows): ISK 800K/month retainer + ISK 75–110K/month seats (28–38 seats). Monthly ongoing: ISK 875–910K (approximately €5,850–6,050/month). Onboarding: ISK 840–1,140K. Committed time: 8–12 hours/week.
The customer can explain this to their accountant in one sentence: “We paid a one-time setup fee, and now we pay Sókrates a monthly fee for our AI department, plus the AI tool seats at cost.” No confusopoly. No opaque consumption metrics. No surprise invoices. No “you’ve used too many automations” upsell conversations.
Revenue stream dominance: The retainer is 85–90% of total billings. The seat passthrough is 10–15%. The retainer is where margin lives. The passthrough is a convenience service that simplifies the customer’s vendor management (one invoice instead of managing their own Anthropic relationship).
5.2 Unit Economics
Unit economics are presented per customer company, not per user — consistent with the pricing model.
At steady state (V1, 20+ customers):
| Metric | Value | Basis |
|---|---|---|
| Average Revenue Per Customer (ARPC) | ISK 650K/month (ISK 7.8M/year) | Blended average across company sizes (majority of target segment falls in the 35–55 employee range). Excludes one-time onboarding fee. |
| COGS per customer | ISK 155K/month | Claude Teams seats (~ISK 70K) + Sókrates Agent API inference (~ISK 15K) + founder/team time allocation (~ISK 70K at 20+ customers with basis-accelerated onboarding). |
| Gross Margin | 76% | (650K - 155K) / 650K |
| Customer Acquisition Cost (CAC) | ISK 300K | 1–2 months of founder time for relationship, trial, and conversion. Near-zero for EDIH-IS subsidized trials. Note: onboarding fee covers onboarding costs and hardware — CAC is pure acquisition, not delivery. |
| Customer Lifetime Value (LTV) | ISK 23.4M | ARPC × 36 months (conservative 3-year retention assumption based on tenure-driven value increase — the Sókrates Agent gets more valuable the longer it runs, and the exit option means customers who stay are choosing to, not locked in) |
| LTV:CAC Ratio | 78:1 | Unusually high because CAC is relationship-driven, not marketing-driven, in a small market — and onboarding costs are now covered by the onboarding fee rather than absorbed into CAC |
| Payback Period | < 1 month | First retainer payment exceeds CAC |
How these evolve:
At launch (1–5 customers): COGS per customer is higher because the founder’s time is concentrated. Effective gross margin is lower (~50–60%) but CAC is near-zero for EDIH-IS trials. The economics work because the founder’s time is the only cost and there is no team to carry.
At scale (30+ customers): COGS per customer drops as the basis compresses onboarding time and the plugin library provides reusable components. A team of 3–5 can manage 30–50 customers. Fleet management tooling (monitoring all customer boxes) reduces per-customer operational overhead. Hardware costs decline with bulk procurement. Gross margin converges toward 76–80%. CAC remains low because the Icelandic market is referral-driven — each satisfied customer generates warm introductions. No paid acquisition channel is needed for the Iceland phase.
The critical margin observation: Unlike API-based AI businesses that face structural margin pressure from compute costs (50–60% gross margins), Sókrates’s COGS is dominated by human time, not inference cost. Claude Teams seats are a fixed, predictable cost per customer. The Sókrates Agent’s API inference cost (Claude calls for reasoning) adds a variable but modest component — the agent reasons periodically, not continuously, and the basis reduces redundant reasoning by providing pre-validated principles. As the basis and Sókrates Agent reduce per-customer time investment, gross margin expands with scale rather than compressing. This is the managed services margin structure, not the SaaS margin structure — and for this business model, it is the correct one.
The exit option as retention economics: Customers retain full ownership of their infrastructure at cancellation (see §4.1 for the exit architecture). What they lose is the Sókrates Agent. This means retention is driven entirely by perceived value, not switching costs. The upside: customers who stay are pure signal. The structural advantage: every month a customer stays, their Eidos gets richer and the Sókrates Agent gets sharper at understanding their specific operations, which makes the Sókrates Agent more valuable the longer the engagement runs. Retention improves with tenure, which is the opposite of the typical SaaS decay curve.
5.3 Revenue Projections Summary
Year 1: ISK 120–160M (€800K–1.1M)
30–40 customers by year-end. First customers acquired through the FinTech cluster and EDIH-IS pathways (see §6.1). Referral cascade through the Icelandic business network drives acquisition through the year. Average ACV ISK 8M, reflecting the target segment’s typical company size (35–55 employees). Revenue ramps through the year as customers are acquired sequentially — Year 1 recognized revenue is lower than annualized run rate. Onboarding fees add ISK 15–25M in one-time revenue across the cohort.
Year 2: ISK 350–550M (€2.3–3.7M)
60–80 customers. Expansion revenue from existing customers growing (new hires, new departments requesting Sókrates Agent access, expanded MCP connector scope). Nordic expansion begins (see §6.3). Team growth (see §7.1) enables parallel onboarding across markets. Fleet management daemon monitors all customer boxes, reducing per-customer operational overhead. Average ACV grows to ISK 9–10M as customer organizations grow and engagement scope deepens. Net revenue retention target: 115%+ (expansion exceeds any churn).
Year 3: ISK 700M–1.2B (€4.7–8M)
100–150 customers across Iceland and 2–3 Nordic markets. Compound bundle launched, creating upsell path for existing customers. Average ACV ISK 10–12M. The basis is now deep enough that onboarding in established verticals is a near-automated process, and founder bandwidth shifts to new market development and Compound bundle delivery.
Key growth drivers: Referral density in the Icelandic market (see §6.2), expansion revenue from deepening integration within existing customers, basis-driven onboarding acceleration, and the physical box as a sales accelerator.
Key assumptions: No customer churn in Year 1 (high-touch, embedded switching costs, tenure-driven value compounding per §5.2). 10% annual gross churn from Year 2 (offset by expansion revenue). Claude Teams pricing remains stable. Sókrates Agent API inference costs remain modest relative to retainer revenue (<5% of ARPC). EDIH-IS funding remains available for initial trials (covering hardware and implementation costs). On-premises hardware costs remain in the $300–500 range per unit; bulk procurement reduces this at fleet scale.
5.4 Pricing Strategy & Sensitivity
Price sensitivity is low in this segment for this product category.
The Hagstofa data confirms price sensitivity is low in this segment (see §2.4 Finding 1). The buyer’s primary concern is “can you actually do this?” not “how much does it cost?” The purchase decision is closer to hiring a fractional executive than to buying software — the evaluation is trust and capability, not feature-price comparison.
Upward price sensitivity: If retainers were 20% higher across the board (ISK 480–960K/month depending on size), the total monthly cost remains well below the cost of the hiring alternative (see §3.3). Conversion would likely be unaffected for buyers who have already decided they need external AI capability. The risk is on the margin — companies considering “do nothing” who need a lower entry point to justify the experiment.
Downward price sensitivity: A 20% price reduction (smallest engagement at ISK 320–365K/month) would not significantly increase conversion volume because the constraint is awareness and trust, not price. It would reduce margin without expanding the addressable market. The EDIH-IS subsidized trial already provides a zero-cost entry path for price-sensitive prospects.
The pricing lever that matters is not the retainer level — it is the value delta the customer experiences. A customer whose Sókrates Agent has automated six workflows in the first three months, saving their operations manager ten hours per week, is not evaluating the retainer against a spreadsheet. They are evaluating it against the felt experience of things working better than they did before. The retainer is stable and predictable; the value compounds (see §5.2 for the tenure-driven retention dynamic).
The Sókrates Agent amplifies this dynamic. Because it proactively discovers workflow friction — rather than waiting for employees to request help — the customer continuously experiences new value without any action on their part. The Sókrates Agent doesn’t generate upsell conversations; it generates gratitude. The expansion path, when it exists, is natural: “We’ve added 15 employees and opened a new department — can you extend the engagement scope?” That’s a company-growth conversation, not a feature-upgrade conversation.
Competitive pricing position: Sókrates at ISK 435–910K/month total ongoing cost sits between managed IT services (ISK 1.6–2.5M/month for full IT outsourcing) and fractional CFO/CTO engagements (ISK 400–750K/month). It is dramatically cheaper than hiring internally (ISK 1–1.6M/month for a single AI specialist before overhead). It is more expensive than standalone Copilot licenses (ISK 4,200/user/month) but delivers a categorically different value — an embedded, proactive AI department that works 24/7, rather than a generic productivity tool that waits to be asked. The Lagaviti comparison is illustrative: their per-seat legal AI pricing at ISK 52–95K/seat/month means a 10-person deployment costs ISK 520–950K/month for a single-vertical tool. Sókrates covers all knowledge-worker verticals within the company, with unlimited automations and a proactive agent that never sleeps, for a comparable total.