# How AI agents are shaping the future of work > Source: > Published: 2026-07-14 10:01:00+00:00 I attended several major technology conferences in 2025 where the first AI agents embedded in enterprise SaaS platforms were announced. Some of these agents showed promise and a glimpse into the future of work, while others looked like natural language extensions of a platform’s existing functionality. At the end of 2025, Anthropic and OpenAI launched new AI models and code-generating capabilities. More developers tried [vibe coding](https://www.infoworld.com/article/4058076/vibe-coding-and-the-future-of-software-development.html), and some platforms launched [spec-driven development capabilities](https://www.infoworld.com/article/4166817/vibe-coding-or-spec-driven-development.html). By February 2026, even The New York Times reported that [the AI disruption had arrived](https://www.nytimes.com/2026/02/18/opinion/ai-software.html), noting that code generators were building “apps that may be flawed, but credible.” Wall Street investors took notice of the code-generation improvements and other disruptive factors, driving a selloff in SaaS stocks, now referred to as the “[SaaSpocalypse](https://www.bloomberg.com/news/articles/2026-02-03/-get-me-out-traders-dump-software-stocks-as-ai-fears-take-hold).” Part of their concern stemmed from the belief that CIOs would use AI to [write software that would replace SaaS solutions](https://www.cio.com/article/4148303/cios-rethink-softwares-future-as-ai-agents-advance.html). But I thought differently and wrote a response in my article asking whether [AI is the end of SaaS as we know it](https://www.cio.com/article/4146669/is-ai-the-end-of-saas-as-we-know-it.html). CIOs might use AI to accelerate application modernization, but I doubt they would replace their ERP, CRM, and even smaller SaaS point solutions by building them. Instead, I believed it would be SaaS companies that would take the most advantage of AI code-generation capabilities. This hypothesis drove me to attend nine conferences this spring to see how SaaS companies were launching AI agents and defining a new future of work. I wrote eight articles on [what CIOs need to know](https://drive.starcio.com/cios-need-to-know) about data management, agile organizations, marketing, ERPs, critical process management, and other evolutions to plan for in the AI era. Now, looking across all nine conferences, I can draw some conclusions about how AI agents are shaping the future of work. Here are my learnings and what CIOs need to consider when evaluating and deploying AI agents in the workplace. SaaS companies have very distinct perspectives on the future of work, including the extent to which humans will play which roles and whether and how quickly we’ll see agentic, fully automated work. For example, Atlassian proclaimed, “[step into the future of human-AI collaboration](https://www.atlassian.com/company/events),” while SAP unveiled “[the autonomous enterprise](https://news.sap.com/2026/05/sap-sapphire-sap-unveils-autonomous-enterprise/).” Snowflake aimed to “[make AI real for business](https://www.snowflake.com/en/summit/),” while Appian targeted “[serious AI built on process](https://www.appianworld.com/).” These vendors’ marketers had to decide whether to lead with AI, people, or business in their messaging, but so must CIOs as they contemplate their AI strategies and how to get employees to fully adopt AI agents. Some CIOs see a fully automated agentic AI as the future, with human-in-the-middle as a transitional phase as departments build trust in AI agents’ decision-making and automation capabilities. Other CIOs see AI more as a tool that delivers productivity improvements by augmenting human decision-making capabilities. Many of these CIOs see human augmentation as essential to supporting critical thinking, innovation, and creativity. [Deloitte’s State of AI Report](https://www.deloitte.com/us/en/what-we-do/capabilities/applied-artificial-intelligence/content/state-of-ai-in-the-enterprise.html), published in January, provides a benchmark. It states that 36% of IT leaders expect at least 10% of their jobs to be fully automated in the next year, and 82% expect to reach that benchmark in three years. Many organizations will have a mix of AI agents, choosing automation where reliability at scale is possible, but opting for human augmentation in operationally critical or customer-facing domains. But how CIOs position AI agents is not only an operational strategy; it’s also a cultural statement that shapes employees’ embrace of AI and whether [detractors vocalize job-loss fears](https://drive.starcio.com/2026/03/ai-leadership-job-at-risk-or-career-opportunity/). In the short term, it will also weigh in on which AI agents to use from different partners and which areas to build in-house. Many solution providers are demonstrating significantly more AI agents this year. For example, SAP went from [40 Joule Agents in 2025 to over 200 in 2026.](https://drive.starcio.com/2026/05/autonomous-enterprise-ai-cios/) Three technology capabilities are fueling this significant growth: The result is that [CIOs will have many options about which agents to test](https://drive.starcio.com/2025/10/ai-agents-definitive-guide-saas-security-titans/), but will have to dedicate analysts to understand the capability, cost, and compliance trade-offs. Additionally, expect AI agent capabilities to evolve significantly over the next few years, so CIOs should continuously revisit their decisions regarding deployed AI agents, focusing on performance, benefits, and ROI. CIOs should also watch for signs of [shadow AI](https://www.cio.com/article/1247890/7-steps-for-turning-shadow-it-into-a-competitive-edge.html) and employee confusion about which AI agents to experiment with on different platforms. The AI strategy should include a transparent, defined process for selecting, reviewing, evaluating, procuring, deploying, driving adoption, monitoring, and collecting end-user feedback around AI agents. The apparent ease-of-use of AI code generators may lead some engineering teams to [build AI agents rather than buy them](https://www.cio.com/article/4097339/your-next-big-ai-decision-isnt-build-vs-buy-its-how-to-combine-the-two.html) from SaaS providers. But CIOs should quickly realize that coding is just one step in developing AI agents, and that aggressively pursuing a build strategy can lead to [AI debt](https://www.cio.com/article/4178324/7-sources-of-ai-debt-and-how-to-avoid-them.html) and [increased AI costs](https://www.cio.com/article/4107377/cios-will-underestimate-ai-infrastructure-costs-by-30.html). DevOps teams can code AI agents using tools such as Claude, Codex, Lovable, and Replit — a do-it-yourself approach. Some SaaS companies are providing an alternative, with AI agent development tools that leverage the data, infrastructure, and governance baked into their platforms. Many of these development tools offer flexibility, allowing developer teams to select AI models and development environments. Examples of new and enhanced AI development tools I saw at conferences this quarter include: I also reviewed [Nutanix Agentic AI](https://www.nutanix.com/solutions/ai), a platform-as-a-service for accelerating the deployment of agentic AI workloads, and [Adobe Firefly AI Assistant](https://www.adobe.com/products/firefly/features/ai-assistant.html) for creatives. These development tools can target different audiences. Some look like low-code development tools targeted at software developers, whereas others are [no-code and enable citizen developers](https://drive.starcio.com/2026/05/low-code-in-the-ai-era-cios-need-to-know/), i.e., businesspeople, to [develop applications and agents](https://www.cio.com/article/4176062/cios-are-enlisting-business-users-to-vibe-code-their-own-apps.html). Additionally, some of these tools support spec-driven development and generate artifacts such as product requirement documents (PRDs), data models, and testing capabilities. Before commissioning AI development for apps and agents, CIOs should sponsor proofs of technical, data, modeling, security, and governance capabilities. Between AI agents and the enterprise’s intelligence, including structured data sources, defined business processes, and agent interactions (both human-to-agent and agent-to-agent), lies an evolving “context layer.” This layer refers to the enterprise knowledge that AI agents draw on when evaluating signals and recommending or taking actions. Context may include a knowledge graph, a semantic layer, cleansed document repositories, and other knowledge bases. The context layer, skills, tools, out-of-the-box agents, and governance capabilities are some areas to review where solution providers differentiate. Some examples: CIOs should recognize that while solution providers will compete on capabilities, the real “secret sauce” of the context layer lies in the company’s trusted data, well-defined business processes, and employee adoption of AI agents. AI agents use the context layer, but also tap into skills, which encode the procedures they can follow, and tools, which prescribe the actions they can take. Before AI agents are ready to pilot, their governance, including permissions, approval gates, and other guardrails, must be defined. Other capabilities to look for when defining AI agents include orchestration, testing evals, and observability. In 2025, many solution providers bolted on AI agents to their existing user experiences. This year, many solution providers showcased new conversational user experiences that employees can use instead of traditional ones built with forms, flows, reports, and static dashboards. Conversational user experiences are where AI agents and people come together, whether it’s human-in-the-middle or human augmentation. Solution providers also grouped their AI agents into assistants or coworkers. For example, [Adobe CX Coworker](https://business.adobe.com/products/cx-enterprise-coworker.html) illustrates human augmentation, helping marketers manage campaigns with prompts and monitor their performance. SAP launched [Joule Assistants](https://www.sap.com/products/artificial-intelligence/ai-assistant.html) across several business functions, including finance, human capital, supply chain, and customer experience. Other assistants, such as [Appian AI Copilot](https://docs.appian.com/suite/help/26.5/appian-ai-copilot.html), [Atlassian Rovo](https://www.atlassian.com/software/rovo), [Cisco AI Assistant](https://www.cisco.com/site/us/en/solutions/artificial-intelligence/ai-assistant/index.html), [Nutanix NIVA](https://www.nutanix.com/blog/nutanix-intelligent-virtual-agent), and [Snowflake CoWork](https://www.snowflake.com/en/product/snowflake-cowork/), offer AI-first user experiences to assist different end-user types. CIOs should demo these [AI-first user experiences](https://www.infoworld.com/article/4178415/what-will-ai-first-ux-look-like.html) to glimpse the future of work. Developers are already getting used to these experiences through code generators and vibe coding tools. Now, similar capabilities are being tailored across all business functions. CIOs should ramp up their [change management programs](https://www.cio.com/article/4082282/preparing-your-workforce-for-ai-agents-a-change-management-guide.html) to accelerate the adoption of these AI capabilities. Solution providers are showcasing AI capabilities that can help CIOs [reshape their businesses](https://drive.starcio.com/2026/04/ai-reshaping-business-not-digital-transformation-yet/). But in Q2, there were only a few examples of how AI can help CIOs drive growth, evolve business models, or embed AI into customer-facing products. I expect to see a wave of further AI innovations that will go beyond productivity improvements and efficiencies and help CIOs pursue [growth-driving digital transformation strategies](https://drive.starcio.com/2025/02/cios-drive-genai-digital-transformation/). *Sacolick travelled to conferences mentioned in this article as a guest of Adobe, Appian, Atlassian, Domo, Nutanix, SAP, and Snowflake. In addition, he was hired by Quickbase to speak at its conference.*