Stratenity POV, Technology & Software Industry Scan

The end of growth-at-all-costs, and the arrival of the inference bill

Software spent a decade optimizing for growth with cheap capital and forgiving investors. That era ended. With higher rates and a repriced funding market, the market now rewards efficient growth: the Rule of 40 (revenue growth rate plus profit margin should exceed 40) has gone from a nice-to-have to a survival test, and companies burning cash for growth they cannot sustain are being marked down hard. The core challenge is that the growth playbook that built these businesses is no longer financeable.

Layered on top is a new structural cost. For years software enjoyed near-zero marginal cost per user, which is what justified its valuation premium. AI has changed that. Inference costs scale directly with usage, and a product that runs large-model calls on every interaction can see its gross margin fall from the classic 75 to 85 percent SaaS range toward 50 to 60 percent if the cost is not engineered down. The defining tension of this cycle is delivering AI-powered value that customers will pay for while defending the gross margin that makes software a great business in the first place.

The unit-economics scorecard every board now demands

SaaS financial health is measured by a tight set of ratios, and the market has tightened its expectations on each. The table below sets the benchmarks that separate a fundable business from a distressed one, and the lever that improves each.

The most important shift is that net revenue retention has become the master metric. A business with 120 percent NRR grows 20 percent from its existing base before adding a single new logo, which is why expansion economics now dominate valuation. The math is stark: at 90 percent NRR you are running up a down escalator, spending CAC just to stand still, while at 120 percent you compound. Retention and expansion, not top-of-funnel volume, are where the durable value is built.

Engineering leverage, not headcount, is the new scaling law

Software teams over-hired in the growth era and spent the correction cutting. The lesson is that revenue per employee, not raw headcount, is the measure of a well-run software company, with strong operators pushing $200,000 to $400,000-plus per full-time employee. The frontier now is engineering leverage: AI-assisted development, platform investment, and ruthless prioritization that lets a smaller team ship more.

• Raise output per engineer through AI coding assistance and platform tooling, treating developer productivity as a measurable, investable asset.
• Protect and grow the roles that touch retention (customer success, solutions, product) since NRR is now the dominant value driver.
• Build AI and data skills in-house rather than relying entirely on vendors, because the differentiation is increasingly in how you apply models to your domain.
• Keep spans of control and management layers lean, since the correction proved that org bloat directly suppresses the Rule of 40.

The build now has to be cost-aware and model-aware

Technical readiness in software used to mean scalability and uptime. It still does, but two new dimensions dominate. First, cloud and inference cost engineering: FinOps discipline, right-sizing, and choosing the right model for each task (a small model for routine calls, a large one only where it earns its cost) is now a gross-margin lever, not a back-office concern. Second, model-readiness: whether the company has the data, evaluation harnesses, and guardrails to deploy AI features that are accurate and safe rather than impressive in a demo and unreliable in production.

• A FinOps capability that attributes cloud and inference cost to product and customer, so margin erosion is visible before it hits the P&L.
• An evaluation and testing harness for AI features, because shipping a model that hallucinates in a customer workflow is a churn event, not a feature.
• Clean, governed proprietary data, since in an AI market the durable moat is your domain data, not the base model everyone can access.
• Architecture that lets you swap models and route requests, avoiding lock-in to a single provider's pricing and roadmap.

Security, privacy, and now AI regulation gate enterprise revenue

For B2B software, compliance is not overhead, it is a prerequisite for revenue. Enterprise buyers require SOC 2 Type II attestation before they sign, and increasingly ISO 27001. Privacy law is global and enforceable: the GDPR carries fines up to 4 percent of global revenue, the CCPA/CPRA governs US consumer data, and data-residency rules shape where you can host. Payment features pull in PCI DSS, and health or financial verticals add HIPAA and SOX obligations. The newest and most consequential layer is AI regulation: the EU AI Act, phasing in through 2025 and 2026, imposes risk-tiered obligations with penalties up to 7 percent of global turnover for prohibited uses, and it applies to any company serving EU users regardless of where it is based.

The governance imperative is that AI features now inherit both the privacy and the AI-specific regulatory burden. Every AI-driven output that reaches a customer should carry provenance: what data trained or grounded it, what model produced it, what version of the prompt and logic, and who is accountable. This is not just good practice, it is what the EU AI Act's transparency and documentation requirements will demand, and what enterprise procurement is already starting to ask for in security reviews.

Reliability and trustworthy AI, not feature count, drive retention

In software, customer outcomes translate directly into the NRR that determines valuation. Reliability is foundational: enterprise SLAs demand 99.9 percent uptime or better, and every hour of downtime erodes the trust that renewals depend on. But the new frontier is AI trustworthiness. A copilot or automation feature that is wrong in a way the customer cannot detect is worse than no feature, because it silently erodes confidence and surfaces as churn at renewal.

• Uptime and performance against SLA, since reliability failures are the fastest route to gross-revenue-retention loss.
• Time-to-value in onboarding, because a customer who does not reach an activation milestone in the first weeks is a churn risk regardless of contract length.
• AI output accuracy and explainability, so customers trust and adopt the features you built rather than quietly ignoring them.
• Expansion signals (usage growth, seat adoption, feature depth) that are the leading indicator of the NRR that drives enterprise value.

Platforms, integrations, and the cloud marketplace as a channel

Software rarely wins alone. Integrations and platform ecosystems determine stickiness: a product embedded in a customer's workflow through deep integrations is far harder to rip out, which directly protects retention. Cloud marketplaces (AWS, Azure, Google Cloud) have become a serious channel, letting buyers purchase against committed cloud spend and shortening sales cycles. Model and infrastructure providers are strategic partners whose pricing and roadmap directly shape your unit economics.

The partnership posture that compounds treats every integration and marketplace listing as a governed, revenue-relevant relationship, not a checkbox. Ecosystem strategy is a make-versus-partner decision at every layer: which capabilities to build, which to integrate, and which to resell. The companies extracting the most value are those that make these choices deliberately and track their contribution to retention and CAC efficiency, rather than accumulating integrations that no one uses.

Make pricing, packaging, and AI-cost decisions governed and explainable

Stratenity's view is that software's next winners will be defined by decision discipline, not raw growth. Pricing, packaging, retention plays, and AI-cost tradeoffs are the decisions that determine NRR, gross margin, and the Rule of 40, yet most companies make them in scattered spreadsheets and board decks with no record of the assumptions or the outcome. Treating each as a typed, versioned artifact with explainable reasoning turns pricing and margin management into a compounding capability.

The path forward is sequenced. Instrument the unit economics honestly first, including true AI cost per customer, so the picture is real. Then govern the highest-leverage decisions, pricing changes, packaging, and model-cost tradeoffs, with explainable logic and a human gate. Then build the AI and compliance provenance that both enterprise buyers and the EU AI Act increasingly demand. Each stage improves the metrics that drive valuation, and each decision leaves a trail that survives a board review, a security audit, and a regulator.

Five moves for the next four quarters

• Build an honest unit-economics model that includes true cloud and inference cost per customer, exposing which segments and features actually make money.
• Make net revenue retention the master metric and reorganize investment toward expansion, onboarding, and success rather than pure top-of-funnel growth.
• Stand up a FinOps and AI-cost discipline that routes each workload to the cheapest sufficient model and attributes cost to product and customer.
• Treat SOC 2, GDPR, and EU AI Act readiness as a revenue enabler, building provenance into AI features before enterprise procurement demands it.
• Rationalize pricing and packaging as a governed, versioned decision, testing changes against elasticity and retention rather than intuition.

Five levers with a metric to hold them accountable

• Expansion and retention program: lift net revenue retention toward 110 percent-plus through land-and-expand pricing and success motions.
• AI-cost engineering and model routing: pull gross margin back toward the 75 percent-plus SaaS range by matching model size to task.
• Onboarding and time-to-value redesign: improve gross revenue retention above 90 percent by getting customers to activation faster.
• Sales-efficiency focus by segment: bring CAC payback under 18 months and LTV to CAC above 3x by concentrating on the best-fit segments.
• Compliance-as-revenue: achieve SOC 2 Type II and EU AI Act readiness to unlock and accelerate enterprise deals gated on it.

Related reading

Put this sector view to work with the cross-cutting Stratenity frameworks.