Artificial intelligence (AI) has been the buzziest technology introduced in the past decade. Following the introduction of widely available generative AI tools, the rate of technology change has been phenomenal—though the rate of process change and human adoption has been less impressive.
At my company, Optimizely, we believe we’re already entering the second wave of AI in the enterprise—an entirely new category of AI usage. Wave 1 lasted a good twelve months, with companies first experimenting with generative AI by tinkering with tools like ChatGPT. Most of the content generated with such tools never made it into teams’ marketing or external channels, but nonetheless, a new era began.
The start of Wave 2 was marked by Microsoft releasing CoPilot earlier this year. Coming from the largest vendor of productivity solutions in the world, CoPilot introduced a massive number of users to the powerful advantages offered by adding AI to the workplace. Critically, it also tests the willingness of enterprises to pay for these enhanced capabilities.
Now, during this second wave, companies are adopting AI and applying it to practical use cases, often as part of existing workflows that are long engrained into the enterprise. Business users are engaging with AI inside everyday tools—Microsoft CoPilot inside of Office tools or Opal inside of Optimizely products, for example.
Regardless of how teams approach this new era of AI usage, the new era has undeniably arrived. Enterprise marketers and technologists would be best served to deeply understand the current capabilities and roadmaps of both categories of solutions.
EMBEDDED AI SOLUTIONS
Embedded AI solutions are applications that have been around long before the advent of AI, but have embedded AI to improve every activity their underlying system did before. For example, a sales person might use AI to create a PowerPoint automatically using Microsoft CoPilot, or a content marketer may generate a first draft of marketing collateral inside of Optimizely’s CMP.
Of course, use cases can extend beyond content generation. AI capabilities, like natural language processing, computer vision, speech recognition, and recommendation systems, can greatly enhance the native capabilities of existing systems. These AI solutions provide more value and convenience to their users by automating tasks, personalizing experiences, and optimizing processes. Ultimately, this can lead to greater efficiency and cost reduction (or greater output and speed to market).
Advantages Of Embedded AI Solutions:
- Embedded AI solutions benefit from having a large and loyal customer base, a well-established brand, and a deep understanding of their domain and use cases.
- Embedded AI tools can leverage existing data and infrastructure to train and deploy their new AI models. By taking advantage of data from thousands of previous and current customers, the training sets available are very real and differentiated.
- These solutions offer a seamless and familiar user experience while adding AI capabilities. Users no longer have to switch back and forth between disparate tools.
- Finally, this approach is more affordable and provides greater cost benefit than having multiple systems.
Challenges Of Embedded AI Solutions:
- Embedded AI solutions face the challenge of integrating with legacy systems, which may require significant changes in architecture, design, and governance.
- These solutions may face competition from AI-only solutions that offer more specialized and differentiated features. Embedded AI solutions may also rely on third-party models, such as OpenAI, versus building on top of custom models.
AI-ONLY SOLUTIONS
AI-only solutions are new tools built with AI in mind as their primary functionality, such as Writer, Jasper, and MidJourney. They offer novel AI-enabled features and functionalities, such as copy generation, conversational agents, and even full video and other complex content creation. These solutions aim to create value propositions that offer alternative, not previously possible, solutions to longstanding challenges.
Advantages Of AI-Only Software:
- AI-only solutions have the advantage of being more agile since they can experiment with new technologies without being constrained by legacy systems or frameworks.
- They leverage the latest advances in AI research and development, as well as the availability of open-source tools and platforms. The solutions usually have their own custom LLMs and work on top of customer training sets, which allows more tailored content generation.
- These solutions focus on privacy and security as part of their initial offerings.
Challenges Of AI-Only Solutions:
- AI-only solutions may lack the credibility and reputation of established solutions, making it challenging to gain trust and adoption from customers and businesses.
- These solutions mostly sit outside the day-to-day workflows of business users. They’ll require a lot of switching between tools, which will result in operational inefficiencies.
- The cost of a native AI solution can be high because it involves custom software development, including designing, coding, testing, and deploying a unique AI model tailored to a specific need.
- Users building on their own models have to deal with the ethical and social implications of their AI features, like bias and transparency.
EVALUATING THE TRADEOFFS
The longevity of AI in the enterprise all depends on how well these solutions can address the unique needs of their users.
As we continue further into Wave 2 of enterprise AI, companies have to make a choice: Do the benefits of AI-only solutions outweigh the inefficiencies and costs of going outside the business users’ workflows?
For a lot of companies, embedded AI solutions will offer genuinely helpful integrations to supercharge productivity, while also keeping users within their existing workflows. This may lead to operational savings, as well as real dollar savings, as redundant tech is weeded out.
As companies make the transition from Wave 1 to Wave 2, they shouldn’t have to go through it alone. We’ve learned that vendors can be an important part of this shift, and that they can also play a role in easing the uncertainty brought about by this rapid change. As we all try to identify an AI approach that’s best for our teams, enterprises and technology vendors alike should lean on one another.
As technology continues to outpace the human ability to adopt it, a good place to start is to meet your employees where they are at today.