Artificial Intelligence (AI) might be the industry’s buzzword of the decade, but is it the gold standard? For some enterprise applications, AI has serious potential. McKinsey & Company reports that deep learning algorithms like ChatGPT could add $2.6 trillion to $4.4 trillion annually across industries when trained with corporate data to perform productivity-improving tasks.
As more large learning models (LLMs) are coming to market, you might find that your enterprise customers are asking how they can automate processes and drive efficiency with AI. While eager to inquire, enterprise decision-makers often hesitate to invest in generative AI for their business upon discovering the significant resources, security considerations, and operational changes required to deploy an effective generative AI tool. While trendy, AI adoption still carries a lot of uncertainty and risk.
AI and the AV Industry
The good news is that our industry is uniquely positioned to help in this new scenario. As a trusted resource and advisor for technology decisions, clients trust AV integrators to offer AI-powered solutions that are ready for prime time and have a meaningful impact. These four critical questions can help integrators adapt their discovery phase to incorporate AI into their AV technology roadmap.
Q1. What are the strengths and limitations of AI?
In response to the increased demand, technology vendors have rushed to develop AI integrations for everything — from truncating transcripts into meeting minutes to brainstorming ideas beyond reality. However, integrators should recognize both the efficacy and limitations of AI in truly enhancing the end-user experience. While generative AI can do a lot in its current state, it’s still best suited as a support tool for enhancing and accelerating processes rather than completely overtaking operations end to end.
The key is to identify discrete tasks that AI can successfully handle. If a human must significantly revise an AI’s output to make it usable, AI will waste more time than it saves. Work with clients to establish clear operational guidelines and oversight for AI tools to avoid these limitation roadblocks. This includes defining responsibilities; establishing handoffs between AI and human operators; allocating maintenance resources and having reasonable expectations. Once these boundaries are established, integrators can introduce effective opportunities to accelerate tasks with AI.
Q2. How do the strengths of AI support the client’s goals?
Properly trained AI can extend enterprise bandwidth by accelerating mundane-yet-time-consuming tasks, like retrieving data to answer questions or inform project timelines; providing employee and customer support; and digesting information to guide brainstorming and decision-making. This kind of implementation improves existing information retrieval processes. AI can also automatically compile information from across sources or customize responses to align with user data permissions.
Once you understand the capabilities of AI, it’s time to bring them to your client. To understand how AI can best support their organization, work with them to understand their unique pain points. Remember, the goal is to improve efficiency in existing systems, not replace processes entirely with AI.
AI functions must align with the interests of the enterprise and the evolving processes, demands and expectations of the business. Deployments should complement and improve the existing employee experience. AI should work across devices and processes, offering a reliable and seamless user experience. If an AI application does not make the job easier for enterprise customers, it’s probably not ready for deployment.
Q3. What training resources are available?
When trained with the right data aggregations, AI assistants can produce qualitative and quantitative-driven outputs that help streamline employee experiences, audience engagement, and objective results that benefit an organization’s defined visual goals and output. That being said, an AI’s responses are only ever as accurate as its dataset. Before launching an AI implementation for an application like sales, customer service or troubleshooting support, a business should have solid and up-to-date documentation of its processes and internal knowledge.
While AI implementation can make operations easier, it’s not an effortless addition to a business model. Deloitte cites managing internal data and processes as one of the most reported obstacles in scaling AI. It is important to inform the client about the continuous investment required for AI and discuss how they plan to maintain it. Before deployment, the business should evaluate whether the AI solutions that client plan to adopt are maintainable. They should then develop a long-term support plan and appoint someone to oversee the ethics and accuracy of the AI system.
Finally, ensure your clients can deliver on processes continuously after deployment, even if their AI tool is down. Integrators can help customers identify reliable, intuitive foundational AI tools that will support their business best, but setting realistic expectations that no AI tool will be free of error or downtime is the key to effective process management and planning.
Q4. What data is acceptable to share?
Data security is top of mind for enterprises, and while the benefits of AI are worth discussing, it’s crucial to educate clients about the potential risks linked with AI.
One risk is the “black box” paradox: the inner workings of AI systems are obscure to humans — this lack of transparency may pose challenges and lead to unforeseen consequences as AI evolves. That being said, sharing data and training is essential to successful AI implementation. Integrators must educate their clients to ensure they understand the risk. This will help clients make more informed decisions regarding adopting and implementing AI technologies.
Security is also a discussion point when evaluating whether an AI tool will operate on premises or in the cloud. On-premises solutions offer several benefits, including that data never leaves the local area network. This is especially important for businesses that deal with highly sensitive data.
Additionally, on-premises solutions offer firewall protection and integration with internal contact/user data, which enables AI to learn identity and permissions. Finally, on-premises solutions provide access to highly secure employee-facing AI chat with responses derived from proprietary company data.
On the other hand, cloud-based solutions offer flexibility, high scalability and accessibility for distributed workforces. However, clients have less control over data distribution than an on-prem solution. Data and encryption keys are stored with third-party providers. This means that if there is downtime, clients may be unable to access their data.
While each option has its benefits and drawbacks, integrators should advocate for establishing a solid security protocol and best practices before sharing data with an AI assistant.
Concluding Thoughts
While the AI landscape fluctuates, these four questions can guide integrators in evaluating and recommending the right AI investments to meet enterprise customer demands. While some clients might be ready for AI, other organizations might need more preparation before taking the leap. Emphasizing genuine operational enhancements over fleeting trends is paramount, guiding clients toward integrations that yield enduring advantages.
Tomer Mann is chief revenue officer at 22Miles.
