Solita summit 2026

AI is changing the whole development process, from business planning and ideation to software delivery and production. In this session, Ville Partanen from DNA, a Finnish telecommunications company, shares a VP perspective based on real-life cases: what has worked, what has gone south, and why product ownership and business context have become even more important in the AI era. 

In this session you will learn: 

• How AI is already reshaping the development process, from idea to production.
• Why speed shifts the bottleneck to product ownership and business decision-making.
• Where VPs and senior leaders should focus next to design better organizations and turn AI experiments into lasting development transformation.

Many companies are experimenting with AI, but relatively few are ready to scale it into reliable business operations.

In this session, Miika Soininen, Chief Digital Officer at Ponsse, shares how the company is moving from individual proofs of concept toward a governed, production-ready AI capability built on trusted data, shared platforms, and in-house competence.

Through practical examples and honest lessons from experiments that both succeeded and failed, the session explores how Ponsse is building its data platform, AI landing zone, and governance model while keeping business value and operational resilience at the center.

The examples include making information in technical manuals easier to find and use, supporting requirements specification with AI, and enabling access to enterprise data through natural language. 

How can a public authority modernise a highly complex Frankenstein IT landscape of legacy systems while maintaining critical business processes, compliance and security? Hear us share the keys we found to unlock and accelerate an IT modernisation that is changing our ability to implement new legislation. 

Key takeaways:

• Start with the business problem, not the technology: AI became relevant because legacy IT was limiting our ability to implement legislation and adapt quickly. (AI-assisted software development)
• Build the foundation first: Modern architecture, APIs and decoupling of legacy systems created the conditions for AI-enabled development. (a new technical foundation)
• Governance enables innovation: Legal clarification, information security, risk management and human oversight allowed us to move faster with confidence. (strong governance)
• AI can connect legislation and technology: Our next step is using AI to turn regulatory text into structured, testable requirements with traceability back to the legal basis. (AI-assisted end-to-end solutions) 

In AI-native product development, the real challenge is often not generating more output, but building enough shared understanding for people to make good decisions together. AI can help maintain context, structure information, and support coordination, but judgement, prioritisation, and decisions still need to stay human.

In this session, we share a practical framing for thinking about where AI helps in product development work and where human responsibility should remain.

Key take-aways:

• Where AI can help product development by structuring context and supporting coordination
• A maturity model for shared context: from individual AI use to a shared value stream
• What still needs human judgement, prioritisation, and ownership

This session explores how to design, deploy, and scale custom AI agents in Databricks to automate business workflows in production. It covers the practical foundations of agent development, including evaluation, governance, CI/CD, and LLMOps, as well as deployment patterns using Databricks Apps. The session also touches on Genie, agent spaces, and how these capabilities can help organizations transform business processes with trusted AI.

What will you learn: 

• How to build custom AI agents in Databricks for real business workflow automation
• Best practices for evaluation, governance, and LLMOps in production environments
• How to use CI/CD and Databricks Apps to deploy and manage agents reliably
• Where Genie and agent spaces fit in the broader Databricks AI ecosystem and business transformation journey

Most people have already tested AI in isolated, single-point solutions: using Copilot on an Excel sheet or chatting with an individual PDF report. While these save time, true competitive advantage emerges only when we move beyond isolated datasets toward a unified, organization-wide situational awareness.

What happens when hard performance metrics are seamlessly combined with context such as seasonality, logistics stockouts, campaign timing, and institutional tacit knowledge? And how can different roles leverage this exact same “Company Brain” through their own distinct lenses?

This session explores how AI fundamentally shifts the paradigm of leadership and expert work when we stop chatting with disconnected datasets and start chatting with the entire company.

What will you learn?

• Beyond “Chat with Your Data”: Why single-point AI experiments are just the appetizer, and what it actually takes to build a true “Company Brain.”
• The context layer: How to combine hard numbers with organizational context without embarking on multi-year data cleanup projects.
• Multiple perspectives: How different functions query the exact same knowledge base through completely role-specific lenses.
• The bottom-line impact: How moving from isolated tools to shared intelligence directly accelerates decision cycles, cuts wasted spend, and creates defensible competitive advantage.

Data governance and data management are not new disciplines, but the conditions around them have changed. As companies try to scale AI, data has become a bottleneck: leaders are prioritizing “a governed, reusable foundation” that connects structured and unstructured data. In this expert conversation, we explore what still matters, what is genuinely new and how to turn governance and management into operational value for business, risk, and technology stakeholders.

Neste’s data and AI journey is now taking the next step: from using AI as an individual productivity tool toward adopting more agentic ways of working in an enterprise data environment. This discussion looks at how Agile Data Agents has progressed from early exploration into practical use, and what is needed next to move from faster individual work toward more reliable, context-driven AI workflows across enterprise data work.

In this session you will learn: 

• How Neste’s ADA journey has evolved from early experimentation into practical use
• What adopting agentic ways of working looks like in an enterprise data context
• Why context, trust, and business understanding are critical in the next phase of AI-driven data work

This session explores how AI-powered automation and agentic solutions are fundamentally reshaping knowledge work in healthcare organisations. By moving from safe experimentation in secure Living Labs to real-world implementation via the Agent Library for Health, we demonstrate how to design AI as a true, reliable collaborator. To ensure these complex shifts succeed, we also introduce the Delivery Management Office (DMO) approach—acting as the “air traffic control” to orchestrate the transformation predictably. Through concrete examples, you will see how embedding AI into core workflows reduces mechanical tasks, empowers employees, and ultimately elevates service quality.

• Accelerate adoption with reusable assets: Learn how to leverage the Agent Library for Health—a collection of domain-specific agents and certified components—to speed up development without dictating workflows.
• Ensure safe, compliant experimentation: Discover how the Living Lab concept enables secure, zero-trust simulation of new AI-enabled processes using realistic but compliant data.
• Orchestrate the transformation: See how the Delivery Management Office (DMO) acts as “air traffic control” to manage change load, align enterprise architecture, and scale AI initiatives predictably.
• Design for the human in the loop: Explore the CollabAI for Health approach to co-designing human-centric workflows that elevate process design and reshape the daily experience of knowledge workers.

What do enterprises most underestimate when they start building an agent estate?

As foundation models converge on parity, the real competition has shifted to the layer that governs, orchestrates, and secures agents across the enterprise. IBM has bet heavily on this with watsonx Orchestrate and Sovereign Core. This dialogue tests that bet against European regulatory reality and the practicalities of enterprise integration.

What if AI could tell you which discounts are actually worth giving to customers? McDonald’s and Solita explore how causal data science and machine learning are used to separate real incremental sales from deals that would have happened anyway — and how to turn that insight into targeted, guest-level decisions at scale. The result: a data-driven approach to discounting that makes every euro of discount budget work harder, creating more value for the business and more value for the guest.

NIB’s data platform has already taken two big leaps — from an on-premises environment to the cloud and Snowflake, and from hand-built pipelines to automated DataOps with Agile Data Engine. This presentation covers the third: Agile Data Agent and Agentic AI now do the mechanical modelling work within a guardrailed, business-approved lifecycle to turbocharge data product development from idea-to-production. Hear about:

• Evolution of data engineering at NIB
• How NIB has combined agentic AI and ADE’s new Agile Data Agent, transforming how data products are delivered
• What the future holds for agentic data engineering

AI is accelerating software development and the biggest challenges no longer lie in writing code. In this session, Fingrid shares why software capability has become a strategic issue for a critical infrastructure operator, how AI and agentic development are changing software delivery.

Key takeaways:

• Why software capability has become a strategic issue for a critical infrastructure operator
• Moving from buy before build towards a buy and build approach
• How AI and agentic development change the focus from coding to software delivery capability

Enterprise AI is stalling on ROI because operational blind spots make it difficult to trust agents with high-stakes decisions and actions at scale. See how the new Celonis Context Model brings together hindsight, insight and foresight to give agents the exact operational context they need to reason correctly and act reliably across your operations. Understand how Celonis serves as the missing piece of the AI stack, allowing you and your partners to safely industrialise AI-driven processes enterprise-wide.

Key takeaways:

• Understand why enterprise AI stalls on ROI
• Learn what Celonis Context Model is and how it works
• Understand how Celonis acts as the missing piece in the AI stack to help you industrialise AI-driven processes enterprise-wide

In this session, we will show how we helped a leading Nordic retailer build and deploy a conversational analytics agent on Snowflake, enabling users to ask business questions in natural language directly in Microsoft Teams. Through a live demo and practical examples, we will explore what it takes to move from a promising AI prototype to a production-ready agent grounded in semantic logic, governed KPIs, and real business context.

What will you learn?

• How a leading Nordic retailer enabled business users to access enterprise data through natural language in Microsoft Teams
• What it takes to make a data agent production-ready, including semantic views, KPI governance, and validated query logic
• What conversational analytics looks like in practice, and why trust matters more than just getting an answer

Everyone has access to the same AI models, and next week there will be a better one. What sets a company apart is its memory: the rules, decisions and know-how of how the business actually works. Today most of it is scattered across shared folders, documents, a few experts’ heads and individual people’s AI chats, where neither the rest of the organisation nor AI agents can use it.

How do you build that memory that both people and AI can rely on, and that gets better every time it’s used? Three teams in different industries set out to solve this, and ended up with the same answer.

What will you learn?

• Why your AI edge depends on your company’s memory, not on the model
• How knowledge from daily work becomes trusted knowledge that people and AI can use
• What three teams in different industries learned putting this into practice
• Where your organisation stands today on the path to a company brain 

As AI becomes a commodity, the real competitive edge won’t come from automation. It will come from organizational reasoning. AI-powered individuals are already working at a whole new level but enterprises haven’t caught up. This session introduces the Generative Twin and envisions how forerunning organisations are unlocking entirely new ways of operating.

Key takeaways

• Individual-level agentic AI has raced ahead of enterprise AI: Some organisations are at the “assisted” level, but nobody fully knows yet what “augmented” or “autonomous” will actually look like. Horizontal Slicing is how you find the specific slices your org can adopt now
• The competitive edge comes from collective organisational knowledge: Capabilities, tech debt, dependencies, strategies and chosen tactics etc. The winners are the organizations that can reason from what they collectively know while competitors are rebuilding context from scratch
• The future is already here, but only for true creatives: Everyone has access to same tools. The leaders and teams who learn to creatively bend AI to their will are the ones who create competitive advantage that is hard to catch up. 

Every layer in the data architecture was originally a workaround for something expensive. Starting with storage, compute and memory, and lately the developer time needed to write the data pipelines.

AI is now reducing the cost of time to develop code rapidly. This session looks at what that means for how we design data platforms and if it is time to think in new ways.  

Discussion points

• Many of the patterns we take for granted – normalization, star schemas, batch windows, were all answers to hardware constraints we no longer have. Today some of them are still used, but for governance reasons instead.
• AI has made most layers quicker to build, but the shape of the architecture looks much the same as it did 25 years ago. Worth asking how much of that is deliberate?
• If writing code becomes cheap, where does the effort move? Semantics, context, data contracts and trust are likely candidates.

AI is rapidly changing what it means to write software, but not whether skilled, judgment-driven engineers are needed. This talk maps out how the developer role is branching into new paths and argues that the fundamentals that made great engineers valuable were never just about the code. Drawing on real findings from building our own agent orchestration tooling RoadCrewAO®, the session offers developers and business leaders a way to think about where developers and engineers want to go next, not a single answer for where they must.

Key takeaways:

• The nature of software engineering work is evolving fast, but the underlying human skills remain as valuable as ever.
• As AI takes on more orchestration and code generation, developers are finding new specializations: going deeper into technical craft or security, becoming architects of AI-driven systems, or building evaluation frameworks for AI-generated code.
• Moving toward product, design, or leadership isn’t a new phenomenon, and it isn’t a retreat from engineering. An engineering background makes that transition more credible, not less.
• There’s no single “right” path forward. The goal is to help developers recognize which direction fits how they want to work, and to trust that their fundamentals travel with them wherever they go.

Confront the critical question of how AI impacts our societal institutions and social structures. We’ll delve into the vital difference between AI that disrupts established practices and AI designed to align with human worlds for institutional resilience and evolution.

AI is generating 741% more code, but only 20% more is shipped. In this session, Arto Santala, who has worked over 30 years as a software architect, looks at why, despite 84% using AI tools and only 29% trusting them, the real challenge is no longer writing code but turning it into reliable, valuable outcomes.

What you will learn / key takeaways:

• Why 741% more code results in only 20% more shipped software
• What the gap between 84% usage and 29% trust reveals about AI in development
• Why code is becoming a byproduct of spec-driven development
• Why AI support does not replace senior-level judgment and delivery focus