Cost is no longer the villain in AI automation
For a long time, investing in AI automation seemed like something only tech giants with bottomless budgets could do. That story has changed in a very real way. A recent study by Jitterbit shows that only 15% of IT leaders still see budget as a significant barrier to adopting artificial intelligence in their processes. This means the vast majority of organizations have already found viable ways to get automation projects off the ground without breaking the bank. More affordable tools, flexible subscription models, and market maturity itself have helped dismantle the idea that AI is synonymous with astronomical investment. Today’s landscape is very different from what we saw two or three years ago, when any proof of concept needed budget approval worthy of a critical infrastructure project.
Bill Conner, Jitterbit’s president and CEO, summed up the moment well by saying the data is clear: the era of AI pilots is over and the era of the agent-driven enterprise has begun. According to him, business leaders have moved past budget concerns and are now focused on the strategic imperative of deploying AI at scale in a secure and successful way. This reflects a mindset shift that goes far beyond corporate talk. It matches what the numbers are already showing in practice.
The results back this turning point. About 78% of AI automation initiatives are delivering moderate to high returns, and the share of projects with negative outcomes or negative ROI is a tiny 2.5%. In other words, the tech is working and creating real value for those who invest. That does not mean every project is a smashing success, but the ratio between gains and losses is so favorable that the debate about whether it is worth investing in AI automation has pretty much lost relevance. The conversation now is all about how to do it in the best possible way, and that is where the challenges that really keep tech teams up at night in 2025 come in.
Security and compliance are the new bottlenecks
With budget stepping out of the spotlight, security has taken center stage in AI automation discussions. According to Jitterbit’s research, 39% of IT leaders name security and compliance as absolute priorities when planning and executing automation projects. It is no exaggeration to say this has become the number one filter when approving any new initiative. When a company puts autonomous agents in charge of tasks that used to rely on people, it is essentially delegating decisions to systems that must be trustworthy, auditable, and protected against vulnerabilities.
Prompt injection attacks, leakage of sensitive data through language models, and unauthorized access to APIs are practical examples of risks that barely existed on security teams’ radar before. Now, every new AI agent has to go through a rigorous review before it gets access to any production environment, and that completely changes the deployment dynamic. How fast a company can move an agent from the lab into real operations depends directly on the maturity of its security processes. Those with well-structured processes gain speed. Those without them end up stalling promising projects in endless review queues.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Regulatory compliance adds an extra layer of complexity to this equation. Different industries have specific rules on how data can be processed, stored, and shared, and AI automation has to respect each of them. In Brazil, LGPD sets clear requirements for handling personal data, and autonomous systems making decisions based on that data must be transparent about their criteria. In Europe, the AI Act is creating risk classifications that will define the level of oversight required for each type of application. Companies that ignore these requirements are not only running legal risks, they are also putting customer and partner trust on the line. That is why compliance and legal teams now have a permanent seat at the table in AI project meetings, something that would have sounded strange not long ago.
The most sensitive point within this security discussion is probably the expanded attack surface. Every autonomous agent connected to internal systems, databases, and third-party tools represents a potential vulnerability. When you scale from one to dozens or hundreds of agents, that surface grows exponentially. Security teams need to rethink their monitoring strategies, create granular access policies for AI agents, and implement real-time anomaly detection mechanisms. It is not enough to reuse the same practices designed for human users, because AI agents operate at an unmatched speed and volume, and any flaw can spread much faster than in a traditional incident.
Accountability is the new deal-breaker in tool selection
Accountability for the actions and decisions made by autonomous systems is arguably the toughest issue organizations are facing right now. Jitterbit’s research makes this crystal clear: for 47% of companies, accountability — a concept that includes security, auditability, traceability, and guardrails — is the decisive factor when choosing AI tools. In the software and tech sector, that number jumps to an impressive 61%. In other words, more than half of tech companies already rank this factor above features, price, or ease of use when deciding which platform to adopt.
When an AI agent makes a mistake, who is on the hook? The team that built the agent? The manager who approved its rollout? The AI platform vendor? This lack of clarity creates a dangerous vacuum, especially in contexts where automated decisions affect customers, partners, or financial processes. More mature companies are building internal accountability frameworks that clearly define the roles of everyone involved in an autonomous agent’s lifecycle, from design to continuous monitoring in production. This governance is not just bureaucracy for its own sake, it is what allows companies to scale with confidence and keep things under control when something inevitably goes off script.
Traceability deserves special attention in this conversation. Knowing exactly why an agent made a given decision, which data it accessed, and which rules it applied is crucial to keeping operations under control. Without that visibility, any incident turns into a black box that is impossible to diagnose. Tools that provide detailed logs, audit trails, and mechanisms to explain agent decisions are naturally gaining preference in the market. It is not about distrusting AI, but about creating the conditions for humans to supervise, correct, and continuously improve the work of autonomous agents.
Organizations plan to scale autonomous agents at a fast pace
The numbers around autonomous agent adoption are striking. According to Jitterbit’s research, organizations currently run an average of 28 autonomous AI agents. Over the next 12 months, they plan to scale to 40 agents, a 43% increase. This growth is not uniform and varies quite a bit by company size. Firms with revenue between 100 million and 499 million pounds plan to jump from 31 to 49 agents. Organizations with revenue above 500 million pounds are aiming for an average of 72 new agents, a 48% increase.
These numbers show we are far past the early stage where a company tested one or two agents in a controlled environment just to see if things worked. We are talking about dozens of agents running simultaneously across different areas of the business, from customer support to logistics, plus financial analysis and vendor management. Each of these agents interacts with systems, makes decisions, and executes tasks with varying degrees of autonomy. Coordinating all of this in a cohesive way demands serious planning and the right infrastructure.
Scalability, however, is not just about copying and pasting an agent that performed well in a pilot. Scaling means dealing with integration across multiple systems, managing dependencies, making sure the infrastructure can handle processing loads, and, above all, maintaining delivery quality as the number of agents grows. Many organizations learn the hard way that jumping from five to fifty autonomous agents brings a completely different set of challenges compared to the initial phase. Latency issues, conflicts between agents accessing the same resources, and observability gaps are just a few of the obstacles that appear once operations hit real scale.
Speeding up time-to-market is the top strategic priority
One interesting data point from the study reveals what is really driving the rush to scale AI automation. The main driver of automation strategies over the next 12 months is speeding up time-to-market for new products and services, cited by 38% of respondents. This factor outranks improving customer experience, mentioned by 35%, and reducing technical debt, cited by 26%.
This makes complete sense when you think about today’s competitive dynamics. Companies that can ship products and features faster grab market share before rivals and create shorter feedback loops with their users. AI automation becomes a powerful lever in this process, eliminating manual bottlenecks in development, testing, integration, and deployment stages. Autonomous agents can automate repetitive tasks that eat up tech teams’ time, freeing professionals to focus on higher-value creative and strategic work.
The fact that technical debt reduction ranks third is also revealing. Many companies have piled up years of legacy systems, fragile integrations, and manual processes that drag down productivity. AI automation offers a real opportunity to tackle this debt systematically, modernizing workflows without necessarily rewriting everything from scratch. Agents can be configured to mediate communication between old systems and new platforms, translate data formats, and automate reconciliations that used to depend on spreadsheets and human intervention.
The path to sustainable AI automation
The path to healthy scalability starts with investing in architecture from day one. That means choosing platforms that support centralized orchestration of agents, implementing monitoring pipelines that provide visibility into the performance of each automation, and creating standardized testing and validation processes before pushing any new agent into production. Companies that treat scalability as an afterthought end up building up technical debt that undermines all the investment they have made so far.
Observability is another crucial pillar. When you have dozens of agents running in parallel, you need dashboards and alerts that show in real time what each one is doing, their success rates, and where bottlenecks are emerging. Without this monitoring layer, operations turn into a guessing game where problems are only detected after they have already had a visible impact on the business or the end user.
This moment also calls for extra attention to team composition. Professionals who understand both the business and the technology are key to making sure AI agents are aligned with organizational goals. There is no point in having a technically flawless agent if it is optimizing the wrong metric or making decisions that are algorithmically sound but misaligned with the company’s strategy. The sweet spot is at the intersection of technical know-how and business vision — that is where the best results happen.
Governing AI automation well matters more than just proving that it works, and this mindset is what separates organizations that will see consistent results from those that will get stuck in endless cycles of fixes and rework. Right now, it is less about rushing to automate everything and more about automating intelligently. The financial barrier has fallen, the results are proven. From here on out, the game is all about maturity, governance, and responsible execution 🚀.
