Article
Turn GenAI and automation into measurable value in your outsourcing deals
Published
16 January 2026
AI and automation are no longer experimental but transforming how technology services are delivered, priced, and scaled. Leading providers report productivity uplifts of 20–50% in software engineering and 15–30% operational savings as AI-driven delivery becomes mainstream. Yet, many outsourcing contracts fail to capture these benefits, leaving efficiency gains with vendors rather than clients.
In 2025, the leading Nordic insurance company, TRYG, announced a strategic partnership with a significantly expanded scope for one of the company’s largest existing IT service providers. The insurance group faced the challenge of managing over 1,300 IT suppliers and operating more than 1,000 applications, many of which were due for significant modernisation or replacement.​ The task of legacy modernisation was not new to the company, nor was the pressure to slash IT budgets to bolster financial performance to remain competitive. What was new, however, was the timing.
The agreement, valued at over 550 million euros, will save the company more than 30% of total IT costs over the contract period, including investments in the transition and an extensive portfolio of transformation initiatives. This was made possible by the existing and planned investment in new technology by the service provider, TCS, which proposed a solution with considerably greater leverage of AI and cloud solutions across the entire IT landscape to augment delivery capability, automate core processes, and elevate customer experience.​
Implement advised TRYG on strategy and transaction execution. And in this article, we shed light on the underlying mechanics of these efficiency gains, and how they can be applied to other outsourcing contracts being negotiated these days. What makes this still more art than science is the collective expectation among market participants that AI and automation will bring about efficiency gains. But exactly how large these will be and when they will materialise remains to be seen.
Harvesting tangible efficiency gains
Agentic AI is becoming the norm, but it must be value‑led. Gartner forecasts that by 2028, about one‑third of enterprise software applications will embed some form of agentic AI, and at least 15% of routine work decisions could be made autonomously. At the same time, it expects more than 40% of agentic AI initiatives to be terminated by 2027 due to unclear business value and rising costs, which only underscores the need for contracts that tie AI deployment to measurable outcomes and savings.1
It is important to state that this is not about ‘AI’ per se; rather, it is about benefits for the client. Contracts should capture business outcomes (cost, quality, speed, resilience, experience) regardless of whether they are delivered via AI, RPA, machine learning, automation, cloud‑native tooling, or process redesign. Contract provisions should remain technology-agnostic and include mechanisms that automatically adjust as efficiency improves from any source.2
Â
A strategic lens: Frame AI‑driven automation into partner models and incentives
Some common gaps in deal structuring risk causing ‘value leakage’. Buyers often sign contracts loosely connected to actual AI-enabled performance, lacking automation roadmaps, transparency dashboards, and shared governance that tie fees to concrete outcomes. To address this, three deal principles should be emphasised:
- Explicit AI-driven value levers linked to commercial incentives
- Baseline- and benchmark-anchored governanceÂ
- Pricing agility covering multi-year transformation milestones
This reinforces the collective market movement towards contracts that evolve as AI scales operationally and financially.
Contracts should specify how cost savings and productivity benefits are shared, with automatic fee reductions tied to automation milestones and measurable changes in unit cost, not just anecdotal claims of ‘AI adoption’ 3
Strategic partnering and sourcing in the AI era must therefore shift from a legacy transactional mindset to partnership frameworks that transparently measure and share benefits and risks. This applies across sourcing towers – from application development and maintenance, through end-user computing and self-service automation, to infrastructure and cloud operations.
Automation impact by layer
End‑user computing (service desk and self‑service)Â
AI‑powered resolution is moving from pilots to mainstream. Cognizant describes programmes achieving up to a 50% reduction in average handling time and around a 70% reduction in human intervention for some processes, while others highlight use cases where generative AI cuts certain task times by up to roughly 50%. Such examples make target ranges like 20–50% ticket automation and substantial speed‑ups a reasonable ambition when contracts incentivise self‑service and virtual agents. 4
Application development and maintenance (AD/AM)Â
Across the applications domain, a range of efficiency enablers are now available for full-scale deployment. Copilot‑style coding and automated test generation have shown the potential to cut certain test cycles and developer effort by double‑digit percentages.
Where legacy modernisation programmes hitherto often ran out of steam before they passed the stage gates of an enterprise IT Portfolio Management process, they are now coming back on the agenda with lower cost and much shorter implementation timelines. Service providers position mainframe and legacy modernisation programmes as capable of delivering around 30–60% IT estate cost reduction, including significant savings from mainframe‑to‑cloud migration and application portfolio rationalisation. Other research describes firms combining generative AI with zero‑based redesign to target cost reductions of up to around 25% – and 10–30% savings on software and maintenance by pruning redundant applications.5
Cognizant reports that around 20% of its code is now AI‑assisted, with developer productivity gains of up to about 34% in some groups, while others estimate that generative AI can raise software engineering productivity by roughly 20–45%. Together with experimental studies on AI coding assistants, this supports target ranges of 20–50% uplift in developer productivity when tools are industrialised and embedded in the SDLC.6
Predictive maintenance and observability-driven engineering can prevent a large share of failures and outages, with case studies reporting reductions of roughly 60–75% in unplanned downtime and infrastructure failures, and in some deployments over 70% fewer critical incidents – and providers are building AI-driven platforms for automated bug detection and root cause analysis to make this repeatable.7, 8
Across the QA tool ecosystem, AI-assisted testing has, in some cases, identified several times as many issues as manual testing alone, prompting providers to develop AI-driven quality platforms and ‘AI testing of AI’ patterns. Evolving QA and risk controls – including test automation, model evaluation, and guardrail frameworks – is essential as AI‑infused applications and operations scale.9
Infrastructure, security, network, data centre, server/storage, and CloudÂ
AI‑driven operations are now proven in production. Service providers report that automated configuration, troubleshooting, and proactive monitoring can deliver around 30–40% cost and efficiency improvements in selected operations environments, with predictive scaling and patch automation further reducing costs and manual work. Gartner expects organisations to adopt AIOps to cut mean time to resolution (MTTR) by up to about 40% and increase operational process automation by around 30% by 2027, reinforcing the business case for contracts that monetise uptime and MTTR rather than headcount.10
With AI‑infused applications and agentic operations, organisations need enterprise‑grade assurance: lifecycle controls, evaluation gates, prompt/response logging, and red‑team capacity are becoming standard. As AI‑driven testing and monitoring tools can uncover many more issues than manual testing alone, contracts should explicitly cover responsibilities for AI risk management, model validation, auditability, and incident.11
Commercial constructs that scale with automation impactÂ
Commercial models must evolve in lockstep with automation penetration. Capturing AI’s cost‑saving potential requires redesigning incentives and budgeting, not just adding tools, and reinvesting a portion of savings into further transformation.
Key constructs include:
- Outcome‑weighted pricing with gain‑share bands and floor/ceiling protections, so both parties participate in upside and are protected from extreme variance.Â
- Milestone‑based savings pass‑through when automation reduces manual hours or FTEs (for example, through auto‑test coverage or auto‑bug fixes) with clearly documented baselines.
- Volume/transaction‑based pricing for tickets, incidents, and deployments with tiered discounts as volumes fall, aligning with automation and self‑service adoption.Â
- An innovation roadmap and re‑opener: providers submit and update automation plans, and pricing is reviewed and reset at defined intervals to reflect realised savings and new capabilities.
- Transparency requirements: quarterly automation metrics, baseline tracking, independent verification, and – where appropriate – reference to external benchmarks.Â
- Benchmarking against automation‑adopting peers to enforce downward pricing adjustments as the market standard for AI‑enabled productivity and cost levels evolves.
Prepare for AI‑led deliveryÂ
Successfully adopting AI-led delivery requires a comprehensive rethink of your operating model, roles, and location strategy to keep pace with rapidly evolving engineering and operational demands.
Role evolutionÂ
Leading engineering and operations models are converging on Full‑Stack Engineers (FSE), Software Development Engineers in Test (SDET), and Site Reliability Engineers (SRE) for end‑to‑end accountability and self‑healing services. Partners are increasingly expected to deploy agents across the SDLC – story, scaffolding, test‑case, and validator agents – plus reliability agents in production, reflecting emerging ‘agentic engineering’ patterns described by both vendors and analysts.
Location strategyÂ
As manual work diminishes, traditional large offshore hubs will gradually give way to smaller, high‑skill automation and cloud‑native centres capable of managing AI‑driven operations, often closer to key business markets. Research shows that high‑value AI work tends to concentrate in talent clusters with strong engineering, data, and product capabilities, which should be reflected in sourcing and location strategies.
GovernanceÂ
Governance must bake in lifecycle controls: model/tool selection (including small language models for specialised tasks), evaluation, cost/performance monitoring, security, and responsible AI review. Analysts and regulators alike emphasise the importance of traceability and human oversight for AI‑enabled decisions, which should be codified into supplier obligations, reporting, and joint steering forums.
What good looks like in 2026Â
The following table summarises indicative ranges of potential savings or efficiency gains from AI and automation across each relevant layer of a typical enterprise IT service stack, reflecting the combined insights from the cited sources:
Calling CIOs and sourcing teamsÂ
Data and best practices from major providers and leading analysts now create a compelling imperative: renewals and re-tenders – in 2026 and beyond – must explicitly embed AI and automation impact metrics and incentives into contracts. This will help close the gap between groundbreaking pilot results and contract commitments. Aligning governance and financial incentives with AI-driven value not only ensures immediate savings but also accelerates transformation while de-risking adoption.
But it is not about AI for its own sake; it is about creating client value. By contracting for outcomes and embedding technology‑agnostic mechanisms that automatically pass savings to you, organisations can safely capture productivity, modernise faster, and improve resilience.
Implement’s Tech Sourcing and Legal teams bring a contracting blueprint – spanning end‑user computing, AD/AM, and infrastructure – that links fees to real efficiency impact, embeds assurance, and evolves your operating model for agentic delivery. The result is enterprise‑grade outcomes, measurable savings, and a partner ecosystem aligned to your value.
References
1. Gartner says over 40% of agentic AI projects will be scrapped by 2027. (2025, June 25). Reuters. https://www.reuters.com/busine...Â
2. Creyghton, E. (2025, October 6). AI ambition meets engineering discipline = success. Cognizant. https://www.cognizant.com/dk/e...Â
3. Cognizant. (2024). Cognizant AI business accelerators for operations. https://www.cognizant.com/en_u...Â
4. Generative AI slashes P&C claim times by 50%, Bain & Company says. (n.d.). Insurance Asia. https://insuranceasia.com/insu...Â
5. Creyghton, E. (2025, October 6). AI ambition meets engineering discipline = success. Cognizant. https://www.cognizant.com/dk/e...Â
6. Fitkin, C. (2025, September 20). Quantifying developer productivity gains from AI tools. MetaCTO. https://www.metacto.com/blogs/...Â
7. The advantages of predictive maintenance with Oxmaint. (2025, August 1). Oxmaint. https://oxmaint.com/article/pr...Â
8. Rafalski, K. (2025). How AI predictive maintenance cuts infrastructure failures by 73%. Netguru Blog. https://www.netguru.com/blog/a...Â
9. Reznik, D. (2025, October 9). The big problem with manual testing (and how AI can fix it) (Updated version). OwlityAI. https://owlity.ai/articles/the...
10. Insights from Gartner IOCS 2024: How AIOps and GenAI are revolutionizing IT operations. (2025, January 30). amasol. https://www.amasol.de/en/blog/...
11. Reznik, D. (2025, October 9). The big problem with manual testing (and how AI can fix it) (Updated version). OwlityAI. https://owlity.ai/articles/the...
