By Priyadarshini L, Managing Director and CRO, sa.global India
Every software services leader I speak to has an AI story. A pilot that worked. A tool the team loves. A proposal that went out in half the time. Almost none of them can tell me what it did to their margin.
That gap is not anecdotal. The SPI Research Impact of AI on Professional Services study, which surveyed 146 firms globally including software and SaaS professional services teams, puts a number on it: 58% of firms identify time savings as their primary return from AI. Fewer than 5% cite revenue generation. The investment is real but the business impact is just not keeping pace. And the distance between firms that are racing ahead and those that are not is growing faster than most leadership teams have registered.
Where AI is generating returns today
The current wave of AI adoption is delivering real value in a specific band of the work – and being clear about where that band sits matters more than celebrating the wins.
Proposal and SOW development is where returns are most visible. AI compresses pursuit timelines and raises first-draft quality, which frees senior practitioners for the commercial conversations that win work. Project deliverable creation, go-to-market research, meeting summarization, and contract review generate consistent gains for similar reasons: the inputs are structured, the workflows are repeatable, and AI does not need to reason across systems to produce something useful.
For firms that have not yet captured these gains, the opportunity is immediate. Start here.
The harder truth is that every one of these use cases operates at the surface of the business. None of them touches the decisions that drive margin: how work is priced, how resources are allocated, how delivery risk is identified before it becomes a write-off. The SPI data confirms this with precision – project profitability analysis, demand planning, and resource management all score below the neutral midpoint on AI impact. The tools are in the building. They are not in the engine room.
The performance gap is not closing on its own
The SPI study isolates the top 18.5% of firms by utilization, margin, delivery performance, and profitability. Their EBITDA sits at 24.6%, against 12.3% for the rest of the field. That spread has widened year on year.
What separates them is not more sophisticated tooling. High performers have over 53% of their employees proficient in AI, against 36% elsewhere. They use AI in nearly a quarter more of their projects. Their data quality concerns are lower, their security posture is stronger, and their leadership alignment is tighter. They built the conditions for AI to work before they scaled the investment.
The most instructive signal in the data is what high performers are not doing. They are 40% less likely to have client-facing agentic AI offerings than the rest of the market. They are not productizing AI externally while the internal foundations are still soft. They are generating a higher share of current revenue from AI initiatives despite running fewer external AI products. Internal discipline is outperforming external ambition, and the financial results reflect it.
Software services leaders measuring AI progress by pilot volume or tool count are measuring the wrong things entirely.
The constraint is data, not technology
There is a specific reason why AI is working at the surface of software services businesses and stalling below it.
Knowing if an active engagement is tracking to margin requires live, connected data from the project system, the billing system, the resourcing tool, and the financial forecast. In most software services firms, those systems do not produce a shared operational picture. They produce four separate answers to the same question. Someone reconciles them every Monday morning, usually in a spreadsheet, usually too late to change anything.
AI deployed on top of that environment drafts better proposals. It does not tell the delivery director if the margin on a live engagement is at risk before the sprint closes. That answer requires a connected data foundation, and building it is the work most firms are deferring in favour of more visible AI investments.
The SPI data makes the cost of that deferral concrete. Firms embedding AI across more than 80% of their projects report EBITDA of 23.8%. Firms below 20% report 10.2%. That spread is not a function of tool sophistication. It is a function of the operational readiness that made broad AI use possible in the first place.
Three stages, and where most firms are stuck
AI maturity in software services moves through three stages. Understanding which stage a firm is in (and what it takes to move to the next one) is the strategic question most leadership teams are not asking precisely enough.
At the ASSIST stage, AI supports individual productivity. Documents get drafted faster. Meetings get summarized. Hours are saved. Firm-level performance metrics do not materially shift because the gains are localized to individuals rather than embedded in how the business runs. This is where the majority of software services firms sit today.
At the AUTOMATE stage, AI operates across connected operational data and starts influencing delivery outcomes. Fee burn is monitored continuously. Resource conflicts are surfaced before they become commitments. Scope drift is visible while the project is still active enough to respond. This is not a future state – firms with the right data foundation are running here today, and the margin improvement is measurable.
At the ORCHESTRATE stage, AI coordinates actions across the engagement lifecycle rather than just surfacing information. Portfolio resourcing optimizes in real time. Financial forecasts update as delivery evolves. Risk signals trigger workflows rather than notifications. The operating model changes, and the firms that get there first will be competing on a different basis entirely
Moving from assist to automate to orchestrate requires one thing above everything else: connected operational data. There is no route around it.
What leadership teams need to decide now
The research is unambiguous on direction. The firms investing in data infrastructure, workforce proficiency, and embedded AI workflows are separating from those that are not (and the separation is accelerating).
The strategic question for every software services leadership team in 2025 is not which AI tools to add. It is a harder question: does the operational foundation underneath existing AI investments support outcomes beyond time savings? For most firms, it does not. Building it is less visible than launching a new AI product, harder to communicate to a board, and slower to show results. It is also the only investment that turns the current wave of AI adoption into durable performance improvement.
The era of AI as productivity tooling is not ending; it is just revealing itself as the floor, not the ceiling. The firms that recognize that distinction now are the ones that will define what software services performance looks like when the data from the next benchmark lands.
