Key Takeaways

• A new archetype of SI is emerging — built on AI agents as the primary production system, not as a feature layered onto unchanged delivery models.
• Five dimensions separate genuinely AI-native firms from those using AI as a label: how they organize people, structure delivery processes, build proprietary tools, define their portfolio, and price their work.
• The pricing model is the clearest signal: firms that charge on outcomes believe their value is in results. Firms that charge by the hour believe it is in time spent.
• Leadership behavior is a leading indicator; the five dimensions are lagging ones. Watch for leaders who publicly name the cannibalization risk, speak with operational specificity, and restructure job architectures rather than add training programs.
• Disruption is arriving from three directions simultaneously: AI platform companies entering SI territory from above, enterprise teams building more internally, and VC-backed vertical software vendors converting services to products.
• The five dimensions are a practical self-assessment framework — not just a way to evaluate others.

Twelve months ago, the directional shifts that AI would create for the systems integrator market were already visible. What has changed is the pace — and the specificity of what the new model actually looks like in practice. This post is about that specificity.

A new archetype of systems integrator is emerging. These firms weren’t built to do what traditional SIs do more efficiently. They were built on different assumptions: that the fundamental ratio of human work to AI work in delivery has inverted, that pricing should reflect outcomes rather than hours, and that a team with the right architecture can outdeliver one that is simply larger. The past year has brought major advances in the underlying foundation models — from Anthropic, OpenAI, and Google — alongside a rapid maturation of agentic systems that plan, act, and iterate across complex workflows without human involvement at each step. Consumer-facing agentic tools are proliferating faster than enterprise procurement cycles can track, and the open source AI ecosystem has grown at a rate without modern precedent. The firms building against this moment are no longer theoretical. They exist, they have clients, and some of them have the backing of investors who have spent decades betting on exactly this kind of structural shift.

For established SIs of every size, this creates both urgency and opportunity. The firms that move now — restructuring delivery around AI agents, building proprietary platforms, and aligning pricing with outcomes — have a genuine window to lead the next phase of the industry. The five dimensions that define what that looks like are now legible from the evidence.

Five Dimensions That Define AI-Native Delivery

The most reliable way to assess how far a firm has progressed toward AI-native delivery is to look at how it operates. Across people, process, tools, portfolio, and pricing, AI-native firms make observably different decisions than traditional ones — and those differences are measurable. These are the precise points where the traditional model creates structural friction, and where the new model generates competitive separation.

People: Agents as team members, not tools

The most significant people-model shift in AI-native firms isn’t about headcount size — it’s about how the delivery workforce is composed and organized. In traditional SIs, human capacity is the fundamental unit of production: junior analysts feed work to senior consultants, offshore centers absorb volume, and scaling output means scaling headcount. AI-native firms plan differently. Agents are assigned explicitly to tasks, iterated on, and scaled independently of human hiring cycles. Human roles concentrate on architecture, judgment, and client relationships — not execution volume. The ratio of human to agent work is a deliberate design choice, not a residual.

Soul of the Machine, founded by Sunil Karkera (previously CEO of Designit and BORN Group), operates with approximately seven people and describes AI agents as its “biggest team members.” The company was built around agent capabilities from inception. Its no-POC policy — refusing proof-of-concept engagements entirely and creating working prototypes in the first client meeting, then going straight to production — is a direct expression of this architecture. Karkera’s framing of the underlying shift is precise: “AI has flattened the services model.”

Process: “Behavior-first” delivery over task decomposition

AI-native delivery processes are distinguished not only by speed but by methodology. Where traditional SIs decompose projects into human-assignable tasks, AI-native firms design their processes around agent capabilities from the first step — using agents to analyze, map, and generate work in parallel, building iteratively with each cycle informing the next through AI learning, and structuring engagements so that delivery accelerates over time rather than plateauing.

Mechanical Orchard, founded by Rob Mee (who built Pivotal Labs and sold it to VMware for $2.7 billion), pioneered an approach it calls “behavior-first” modernization — a term that refers not to how people act, but to how software systems actually behave at runtime. The insight is that legacy systems accumulate decades of undocumented business logic: edge cases, workarounds, and institutional knowledge that live in what the code does, not what it says. Traditional SI modernization projects attempt to translate that code line by line, a process that is slow, fragile, and routinely fails to capture the logic that actually matters. Mechanical Orchard’s Imogen platform takes a different route: it observes and maps actual system behavior through live data flows, then uses generative AI to recreate that behavior in modern cloud-native code. The result is a modernization path that starts from what the system does — and reliably reproduces it — rather than from what its code says it does. Gartner validated the approach directly, recommending that software engineering leaders adopt behavior-first methodology. The company reached profitability at more than $10M ARR within 14 months of founding, with roughly 100 employees. Its $50M Series B from GV was unsolicited — a signal of investor conviction that is relatively rare in services businesses.

Tools: Proprietary platforms as the basis for real leverage

How a firm equips itself for delivery is where genuine IP separates from surface differentiation. Using the same foundation models, the same cloud platforms, and the same off-the-shelf agent frameworks as everyone else produces parity, not advantage. The firms building durable positions have invested in proprietary accelerators, platforms, or frameworks that create delivery leverage their competitors cannot replicate by purchasing the same subscriptions.

Qubika’s QBricks platform, launched in early 2026 on Databricks, provides pre-built agents, visual workflow builders, evaluation frameworks, and production-ready agent ecosystems purpose-built for regulated industries. Aivar Innovations built three distinct proprietary platforms from inception: Convogent AI for conversational intelligence, Velogent AI for agentic process automation across invoice processing and compliance workflows, and Kubogent AI for enterprise Kubernetes infrastructure. Mechanical Orchard’s Imogen includes a patented method for producing deterministic outcomes from generative AI — a meaningful technical distinction in an environment where output reliability is a production requirement.

Portfolio: Focused depth over full-service breadth

AI-native firms tend to make sharper portfolio choices than their traditional counterparts. Rather than presenting broad, multi-practice capability maps, the strongest firms stake out specific problems, verticals, or technology domains and build compounding depth there. This is a strategic choice that reflects a different theory of competitive advantage: defensible AI capability requires domain-specific training data, workflow expertise, and accumulated agent intelligence that cannot be replicated quickly. Horizontal coverage at the traditional SI scale requires headcount. Focused depth with AI leverage requires something different.

Qubika illustrates what this looks like at mid-market scale. Backed by Recognize Partners and employing between 500 and 1,000 people, the company has codified three AI delivery pillars — AccelerateAI, Agentic Factory, and Data Foundation — and built named enterprise clients including Walmart’s ONE platform across focused industry verticals: banking, fintech, insurance, healthcare. A PwC-audited NPS of 82 reflects client response. For the many mid-size SIs navigating this transition from an established base rather than a clean-sheet founding, Qubika provides the closest available template.

Pricing: A firm’s pricing model reveals what it actually believes about its own work

How an SI structures its fees is the most direct expression of its underlying business logic. A firm that charges by the hour, regardless of what is delivered, is operating on the premise that its value is in human time spent. A firm that prices on outcomes is operating on the premise that its value is in results produced. The pricing model is not an administrative detail — it is a statement about what the firm believes it is in business to do.

Gruve.ai, founded by Tarun Raisoni (who previously built Rahi Systems to $500M+ revenue before its acquisition by Wesco), uses usage-based, outcome-oriented billing — described as services available “like electricity or cloud computing.” For AI-powered security work, the company charges upon breach detection and analysis, not upon implementation hours. Mayfield partner Navin Chaddha has stated that Gruve targets software-like gross margins of 70–80% — compared to roughly 30% at traditional IT services firms — enabled by AI agent automation. Mayfield is historically averse to services businesses; Chaddha described this as the first time he has seen IT services merit VC economics. That judgment, from a firm with decades of pattern recognition in technology investing, is itself a signal worth noting.

Globant — a NYSE-listed SI with Q4 2025 adjusted gross margins of 37.6% — represents the most significant structural pricing innovation from the established side of the market. Its AI Pods model replaces traditional FTE-based billing with a monthly subscription plus token-metered consumption model. CEO Martín Migoya has framed this explicitly as “services as software,” with a Netflix analogy: enterprises subscribe to a base capability and consume additional capacity through tokens as demand dictates. What Globant demonstrates is that the pricing transition does not require a clean-sheet founding — it is achievable from within an existing SI structure at meaningful scale.

Leadership and Culture: The Signals the Five Dimensions Cannot Capture

The five dimensions above measure what a firm has already done. They are, by definition, lagging indicators: evidence of decisions made, capital deployed, and structures built. They tell you where a firm is. The forward signal — the clearest indication of where a firm is heading — lives in how its leaders engage with the transformation publicly and internally.

Every major shift in SI delivery model visible across those five dimensions began as a leadership conviction before it became a measurable outcome. Accenture’s $5.9 billion in GenAI bookings in FY2025 was the result of decisions Julie Sweet, Chair and Chief Executive Officer, made in 2022 and 2023 — restructuring the firm’s operating model, tying promotion to AI tool usage, and publicly committing to cannibalizing delivery revenue before the market forced the issue. Globant’s token-metered AI Pods pricing model originated from a founding team that had talked openly about AI-driven delivery economics for years before the product existed.

The pattern holds in reverse, and it is worth naming honestly: the firms facing the hardest transformation are those where a large, profitable, deeply embedded delivery model creates structural resistance to exactly the changes the five dimensions require. This is not a leadership failure — it is among the most genuinely difficult strategic challenges in professional services. Firms navigating this transition carry both the weight of existing client commitments and the responsibility of tens of thousands of people whose roles are evolving. The leaders doing this work well are operating under conditions the AI-native boutiques simply do not face. That context matters when reading the signals below — they are not a scorecard, but a set of observable markers that track where conviction is translating into structural change.

There are four such signals worth watching.

Willingness to name the cannibalization risk. The most credible AI-first leaders engage openly with the uncomfortable truth that AI productivity gains compress billable hours under a time-and-materials model. When TCS CEO K. Krithivasan publicly stated that AI gets “first right of refusal” in every engagement — even if it means reducing delivery revenue — that was a CEO naming the cannibalization risk on an earnings call, on the record. That kind of statement matters because it creates public accountability and signals that the internal conversation has already moved past denial. Leaders who use words like “cannibalize,” “structural change,” or “revenue compression” without immediately walking them back are signaling that their mental model has already shifted.

The specificity of the AI narrative. There is a meaningful difference between a leader who describes broad investment in AI and one who can articulate — in operational terms — what their firm’s AI delivery model actually looks like: which agents are in production, what the measured productivity improvement is, and where the model still needs work. Specificity is the distinguishing signal. Leaders who will say “our agent reduces code review time by 60% on engagements above a certain complexity, but we have not yet solved the integration testing problem” are building something real. The precision of the language reflects genuine operational depth.

The hiring signals they send publicly. Job postings are public documents and a reliable proxy for internal strategic commitment. A firm that lists “AI fluency required for all roles” across engineering, delivery, and management is making a different bet than one that lists “AI experience a plus” in a subset of specialist roles. Leaders who restructure job architectures — rather than adding training programs — are making a harder, costlier, and more credible commitment to the delivery model the five dimensions describe.

How they talk about pricing. This is the clearest signal, because pricing is where conviction meets commercial risk. Leaders who articulate publicly — to clients, to analysts — that FTE billing is structurally misaligned with AI-driven delivery are leaders who will eventually be held to that position. When a leader describes outcome-based pricing not as an experiment but as the direction of travel, that is the strongest available forward indicator.

For SI practitioners and leaders reading this: the five dimensions reflect observable evidence from the firms building the new model now. The firms that move most significantly over the next 12 to 24 months will be those where leadership has already begun translating conviction into structural action — naming the cannibalization risk, committing publicly to delivery model innovation, and redesigning job architectures rather than layering training programs onto unchanged roles. The measurable outcomes follow the conviction. Conviction has to come first.

Leadership narrative sets the direction, but culture is the medium through which it travels or stalls. The most telling cultural signal observable from outside a firm is whether the AI narrative flows top-down only, or whether it is also being generated bottom-up by practitioners. A firm where senior leadership talks about AI transformation but engineers and delivery managers describe their day-to-day work in entirely traditional terms has a gap that will eventually show up across all five dimensions. A firm where engineers, delivery leads, and mid-level managers independently publish and debate AI-native delivery approaches — without being prompted by a marketing team — has a cultural reality the measurable dimensions will eventually catch up to. Culture is the hardest thing to fake and, for exactly that reason, one of the most reliable things an outside observer can read.

The Disruption Is Coming From Multiple Directions

The competitive challenges facing traditional SIs are increasingly well-understood in the industry. What is less commonly synthesized is the fact that pressure is arriving simultaneously from at least three distinct directions, each with different characteristics and timelines.

The first is from the AI platform companies themselves. On February 23, 2026, OpenAI launched Frontier Alliances — formal partnerships with Accenture, BCG, Capgemini, and McKinsey to deploy its enterprise AI agent platform — while simultaneously building its own consulting workforce and forward-deployed engineering capacity. AI foundation model companies are becoming de facto systems integrators. The integration layer that traditional SIs have long owned is being commoditized from above by the very companies whose models SIs are being paid to implement.

The second is the reduction in the minimum viable complexity threshold for building internally. As agentic frameworks become more accessible and model capabilities continue to advance, enterprise technology teams are building solutions that would previously have required SI engagement. This is not insourcing in the traditional sense — it is a structural shift in what enterprises can do for themselves on defined AI use cases, accelerating as consumer AI tools demonstrate what is possible and internal teams develop the confidence to replicate it.

The third is the large volume of venture capital flowing into product-based vendors explicitly converting traditional services categories into software. These firms — typically approaching specific industries or workflow categories vertically — are replacing customized SI-delivered solutions with repeatable AI-native products. When the economics of that model mature, they are structurally superior to services delivery, and the capital investment in making them work is substantial and accelerating.

No single one of these pressures is decisive in the near term. The combined effect — with each vector reinforcing the others — is a more complex competitive environment than any prior wave of SI market disruption has produced.

Where to Take This

The five dimensions above are most useful not as a descriptive taxonomy but as a practical self-assessment framework. For any SI firm — of any size, at any stage of this transition — the most productive starting point is to answer each dimension honestly for your own organization:

• What is your actual human-to-agent ratio in a live engagement today, and how has it changed over the past 24 months?
• Have you built proprietary delivery platforms that create leverage your competitors cannot replicate — or are your AI tools the same ones available to any firm with a subscription?
• Have you made real portfolio choices, building compounding depth in specific domains — or does your capability map still cover everything at roughly equal depth?
• What does your current pricing model say about your confidence in AI-assisted delivery — and are you willing to put fees at risk on the outcomes you are promising?

The firms profiled in this post are at early stages of development relative to the scale of the established industry, and the models they are proving have not yet been tested at that scale. But they represent the clearest available reference points for what AI-native delivery looks like in practice across each of the five dimensions. Using them as inputs to your own planning — not as a competitive threat to react to, but as a set of working hypotheses about where the value is being built — is how the most forward-looking firms are already approaching this work.

The Tools dimension is where StackGen shows up most directly in this picture. The proprietary platforms that AI-native SIs are assembling — agent frameworks, automation infrastructure, deployment tooling — are not built in isolation. The firms on this list are constructing them from components, and the infrastructure and automation layer is where we work alongside them. If you are building toward any of the five dimensions above and want to compare notes on the tooling side, we would welcome that conversation.

This is a post  in a multi-part series on AI and the restructuring of the systems integrator market.   Other parts will address the enterprise buyer perspective: how to assess your current SI relationships, what to do if you are falling behind, and how to build for AI velocity without rebuilding your operating model from scratch.

About StackGen

StackGen is the agentic platform for autonomous operations — built for AI-native and AI-first SIs who are constructing the delivery model this article describes, and for the enterprise engineering teams who are their counterparts on the other side of those engagements. We work with firms across the spectrum of this transition, from purpose-built AI-native SIs to traditional firms navigating the shift. The journey described in this article is one we work through every day.