Just when you were starting to get your bearings with generative AI, the conversation has changed again. While generative AI still has a big role to play in business, the newer technology that's reshaping organizations is its successor: agentic AI.
Agentic AI builds on the capabilities of generative AI by connecting advanced AI models to various business tools so they can take direct action. AI agents are provided the autonomy to evaluate information, make decisions, and take actions based on their analysis — all potentially helping companies accomplish more in less time, with less work.
The new question organizations face is how to adapt to this new era. How can you develop the right enterprise AI strategy to embrace the agentic future?
The Shift to the "Agentic Era"
If you're pondering this question right now, you're not alone. As big of a deal as generative AI has been, businesses are investing even more into developing AI agents. In recent research from Omdia, 42% of organizations reported allocating over $1 million to AI agents over the next 12 months, while 80% consider agentic AI a top priority.
When you consider the many areas of business operations where AI agents can contribute, the level of interest in them just makes sense. Across industries, organizations have already started putting agentic AI to good use. To give you an idea of how that looks in practice, here are a few common use cases.
- Finance: In a recent report from Ardent, 41% of finance professionals listed slow invoice and payment approval processes as a top challenge. Using agentic AI to set up automated processing for invoices and POs (purchase orders) can take a lot of tedious work off your finance team's plate, ensuring partners get paid faster, your relationships stay strong, and your finance team has more time for other business-critical tasks.
- Customer Experience: Research from HubSpot found that 82% of customers expect immediate problem resolution from customer service (CS) agents — but your human customer service staff can't be available 24/7. Companies can use Dialogflow to build sophisticated chat assistants that can handle many customer service issues autonomously. That means customers get answers immediately, and your human CS staff have more time to handle complex issues that require a human touch.
- Operations: Your IT department likely also spends a lot of their time on tedious, repetitive tasks. By using AI agents to introduce intelligent ticket routing and business development automation to their processes, you can remove much of the tedious work that fills their days now, while ensuring employees get the help they need faster at the same time.
42% of organizations reported allocating over $1 million to AI agents over the next 12 months.- Omdia
The "Build vs. Buy" Dilemma
Developers now have the tools to (theoretically) build custom AI agents for just about any task you could want. But building an agent that does precisely what you need — including integrating with your other tools and agents, and providing appropriate security and compliance guardrails — requires a good amount of skill and time. 85% of people in Omdia's survey agreed that prebuilt agents enable faster implementation.
That leaves you with a choice. You can evaluate existing AI agents offered by software companies that provide the functionality you need (or something close to it); or you can commit resources to having your own engineers build agents that meet your needs more precisely. 89% of respondents in Omdia's research agree that building an AI agent would require significant resources, and 81% say it involves higher risk and complexity. But 84% agree that it also means more control and flexibility.
In practice, you'll probably want to do a mix of both options: prebuilt agents for cases where something already exists that works for your needs, and custom agents for situations where your needs are more specific. Consider each use case on its own merits, and weigh the differences of building versus buying for that particular function.
Pros & Cons of Prebuilt Agents:
- Faster: An agent that's already built will take less time to get set up and start using. If speed is an important priority, then finding an AI agent that already exists to handle a task is your best shortcut to getting started fast.
- Often More Affordable: A prebuilt agent will cost you money upfront to buy, but often the immediate cost is significantly less than what you would spend on the labor and resources needed to build an AI agent from scratch.
- Less Flexibility: Software companies that build AI agents to sell aren't thinking about your specific needs and priorities, they're building something they hope will work for multiple clients. That means it may meet your primary needs, without matching all your preferences.
- Potential for Easy (if Limited) Integrations: AI agents are often built by vendors to work seamlessly within the products they offer. For example, a Salesforce agent designed to work within your Salesforce ecosystem will have easy access to all the data that lives there. That saves you the work of setting up that integration, but can put a limit on how many integrations the prebuilt agent is good for.
Pros & Cons of Building Customer Agents:
- Specific to Your Needs: This is the big benefit of building your own AI agent. If you can't find a prebuilt one that does what you want, you can create an agent that meets all your particular needs. Where a prebuilt agent may integrate well with one or two other tech products, you can build one designed to work seamlessly with all relevant systems. If you work in an industry where security and compliance are top priorities, your team can make sure to build in all the necessary guardrails.
- Differentiation: When you have easy access to a prebuilt AI agent, so do all your competitors. Building a custom agent that accomplishes something your competitors can't easily do can provide a competitive edge.
- Longer Time to Value: Building a custom agent can take a long time. And the more specific and complex your preferences for the custom agent are, the longer it will take. You may get more value out of it once it's done, but you'll have to commit a long period of time to reaching that point.
- More Expensive: The money you spend on labor and the technical resources required to build your agents will add up. The more powerful your agent needs to be, the more it will cost to build it to your specifications.
Common Agentic AI Challenges to Address
The choice to incorporate agentic AI into your business is an easy decision to make, but actually doing the work to make it happen is the harder part. Organizations typically face a few main challenges to developing and using AI agents effectively.
1. The Scaling Gap
Many organizations eager to embrace agentic AI start with bigger aspirations than they can realistically implement right away. According to Omdia's research, 81% of organizations find scaling AI agents to production more complex than expected. There are a few common reasons for this:
- Compute and Storage Demand: Using AI agents requires high-performance GPUs, a scalable cloud environment, and vector databases. Many organizations quickly realize the amount of compute and storage they need to design and deploy the number of AI agents they want is higher than they expected.
- Lack of Governance and Strategy: The more AI agents you're hoping to create, the more important it is to develop a cohesive strategy to make sure they all work together. Before you start creating disconnected AI agents left and right, you want to take a step back and develop a framework for agent orchestration and monitoring.
- Skills Gap: Agentic AI is new, and many organizations will find their developers just don't have the skills yet to build, launch, and manage the kind of agents you have in mind. You may need to invest in upskilling your current team, hire new staff with the right qualifications, or bring in an outside partner to help.
- Higher Costs than Anticipated: The technology and labor costs associated with scaling AI agents are often higher than organizations understand in advance. Having a realistic idea of expected costs is crucial to creating and executing on a strong enterprise AI strategy.
2. Integration Issues
For AI agents to do the jobs you design them for, they'll generally need access to your data and systems. Most organizations use a large number of disparate systems to handle business data and various business functions. Connecting your AI agents to all the necessary systems in ways that maintain your security is a significant hurdle.
Unsurprisingly, 68% of organizations in Omdia's survey said they've faced significant challenges integrating AI agents into existing systems and processes. Add in how common data silos are in large organizations and how often data is in different formats across different products, and you get a big bottleneck that has to be dealt with before AI agents can do their part.
Data harmonization is one big challenge. Managing multiple integrations with different systems is another. Both can significantly slow down your agentic AI projects.
3. Security Concerns
Agentic AI may have great potential, but it comes with great risk. A SailPoint report found that 96% of technology professionals see AI agents as a growing security threat. That's unsurprising when coupled with the reveal that 80% of organizations in the survey said their AI agents had taken unintended actions — examples included inappropriate data sharing, unauthorized system access, and revealing access credentials.
In Omdia's research, respondents named security and compliance as the number one challenge for implementing agentic AI. Any good enterprise AI strategy must treat security as a top priority. That means thinking through some of the main risks that come with scaling agentic AI, such as:
- Expanded Attack Risk: AI agents designed to interact with multiple third-party systems and the larger internet have significant entry points for attacks. Bad actors can set up malicious prompt injections to try and trick an AI agent into providing classified information or taking dangerous actions.
- Data Breaches: For AI agents to work effectively, they need access to your proprietary data. But the more information they have access to, the more risk there is that they may leak or disclose classified information.
- Too Much Autonomy: One of the biggest benefits AI agents have to offer is one of the biggest risks they bring as well: autonomy. Agents can be designed to make decisions and act on their own, which is good when that means taking tedious work off employees' plates. But if they make the wrong choice — like deleting valuable data or offering customers a 100% discount — it can potentially have big consequences for your business.
- Ethical Issues: 64% of respondents in Omdia's research consider managing AI ethics like fairness, transparency, and bias to be a major obstacle in using them. AI is prone to replicate the biases in its training data, which can lead to issues like inadvertent discrimination.
- Hallucinations: Generative AI products are known for confidently providing incorrect information. Agentic AI can bring similar risks. You don't want AI agents giving your employees or customers answers that are wrong.
- Compliance Concerns: If your industry has stringent compliance requirements, like audit trails and bias monitoring, scaling agentic AI too fast can make it hard to stay on top of those requirements. As you deploy more and more agents, monitoring all of them can become an overwhelming challenge.
What to Look for in a Partner
Navigating all these challenges to effectively build and act on an enterprise AI strategy is no small feat. But that doesn't mean it's not worth doing. One of the best ways to improve your odds of deploying agentic AI successfully is to find the right partner.
Indeed, 91% of organizations told Omdia that the right external partner was critical to their deployment of agentic AI. Finding a partner that has experience with the technology and brings a strategic, future-proof mindset to how they approach agentic AI can help you ensure you get the most out of using AI agents.
Beyond having worked with AI agents before, there are a few key capabilities you should look for in an agentic AI partner:
Data Grounding
To realize their potential, AI agents must be grounded in an organization's proprietary data. The right agentic AI partner should have experience grounding agents in enterprise data, while staying compliant with relevant regulatory requirements like GDPR (General Data Protection Regulation) and HIPAA (Health Insurance Portability and Accountability Act).
Grounded agents are able to perform more complex actions based on extensive data analysis and training. They need a way to evaluate all the information about what's come before to make the best decisions moving forward, which means access to all your relevant data sources. A partner can help you build compliant agents that connect seamlessly to your other software products in order to access the right data.
Change Management
Like any new technology you add to your stack, adopting AI effectively isn't just about the technology itself. Integrating AI agents into your internal processes is just as much about human transformation — making sure your employees understand the technology, embrace it, and learn to incorporate it into their workflows.
The right partner won't just focus on how to build AI agents, they'll also help you develop a strategy for introducing the technological change to your staff. They’ll help you earn buy-in for using AI agents, and train employees how to use it most effectively.
And crucially, they'll understand the different roles AI agents will play in different departments. One overarching change management strategy won't necessarily work here, unless it tailors more specific strategies to each department based on the departmental goals and culture.
Platform Access
The final important capability to look for is expertise with the specific enterprise agentic AI platform you hope to use. Not all AI technology is the same. If your organization uses Google products in most arenas, then Google Cloud’s Gemini Enterprise is the most obvious choice to build your agentic AI strategy around.
Choosing Gemini Enterprise can help proactively address data and integration challenges by bringing in data from over 100 different applications, including common business tools like Salesforce, SharePoint, and Jira.
And as an industry giant well aware of the importance of security and compliance, Google's agentic AI products are built to provide enterprise-grade security options to keep your data safe and your business on the right side of all industry regulations.
Selecting a partner that has extensive experience with Gemini will be your best bet for building an agentic AI strategy that works seamlessly with the technology your teams already use every day.
The Right Agentic AI Partner
For any business using Google products, Promevo is a strong choice for helping you build out your enterprise AI strategy. We can help you identify the most valuable use cases, determine the areas where it makes sense to build agents and when to buy, and help you train your teams in how best to use AI agents within their workflows.
For a more detailed report on how to embrace agentic AI within your enterprise successfully, read the full whitepaper Promevo developed in partnership with Omdia. You'll get a firmer bearing on the current state of the agentic AI market, the challenges involved in effectively implementing AI agents, and how to find the best partner to help you move toward an agentic future.
