How AI bridges the gap between insight and action

Every interaction between your business and your consumers has value somewhere to someone in your organization. This is true no matter what sort of business you run, what channels you use and what customer segments you target.

Most organizations are aware of this, but there has always been a big gap between awareness and adoption. Technology is the enabler, but it has not always been able to capture interactions at scale in a consistent way, nor deliver these insights in a way that supports decision-making—until now.

Artificial intelligence (AI) can do all of the above and more. However, AI is not in and of itself the silver bullet: There are many moving parts that need to be addressed, from how the interactions are sourced and surfaced and from where to what sort of AI you use for what use case.

Here are three perspectives, based on AIntensify’s four-plus years of experience running a live AI-based solution for contact centers, initially in cruise but now applicable to any sector inside and outside travel.

Strategy

Everybody is talking about AI strategies, and there are some companies making a mistake by starting out too big with unrealistic goals.

Differentiating between augmentation and automation is a useful first step. Augmentation is where AI is used to enhance existing workflows and processes, often in sync with a real person. Automation is where these processes are outsourced entirely to an AI [model], which wants to minimize human involvement.

For most travel businesses, augmentation is a more effective entry point. It means that an “AI strategy” can be framed as a “business plan” that identifies use cases within your organization in need of improvement. Sequential improvements through augmentation deliver quick and visible wins if the technology—in this case the AI—can be onboarded in hours rather than weeks. 

Automation can deliver bigger wins over time, although this is not guaranteed, and the direct and indirect costs can be significant.

Our reading of the current AI landscape is that it is better to buy rather than build, to start narrow and go deep.

In our field, AI-enabled voice recognition can process audio files into structured data points, transforming customer interactions with the contact center into pre-defined and customizable fields within a new or existing customer relationship management (CRM) [system] or other relevant systems. The entire process will continue to improve as more structured insights are fed into the database on an ongoing basis. It’s how the machine learning element of AI works.

Signals

Application programming interfaces (APIs) should not be overlooked as a facilitator of AI innovation. APIs make it possible for the data from contact center conversations to be converted from unstructured data into structured data, informing other systems across the business. The connection can be direct to the system or into a centralized database, which can be ingested and then distributed via the AI.

Think of the constantly updated and improved CRM as a multi-channel transmitter, sending signals out to the satellites in a way that supplements the data already contained within the satellite. The net result is that each satellite becomes more intelligent because of the additional signals it receives. And when satellites are already talking to each other, the net benefit is cumulative.

There’s a range of satellites where the signals can have a positive impact. Revenue management teams can flex their pricing strategies and update demand forecasting based on what customers are booking (or not booking). Marketing teams can evaluate campaign performance and refine strategies based on how callers reference (or fail to reference) the latest TV spot or sponsorship deal. Granular and bigger picture feedback into the product itself can inform improvements and innovations.

Signals shared between departments break down silos, foster collaboration and improve overall business performance.

Orchestration 

One advantage of working with a specialist partner rather than trying to do it in-house is that specialists are on top of developments in the AI market. The differences between the large language models are subtle but significant. Identifying which version of which model is best suited on a case-by-case basis needs specialist insight.

The best model to surface pricing insights and send signals to revenue management is not necessarily the best to identify sentiment and send signals to customer service.

This is because there are models for different layers, trained specifically on sentiment, summarization, reasoning and more. There are also models trained on surfacing these layers for specific verticals, including travel.

There is often another layer of orchestration needed, namely understanding the use cases where proprietary AI models can outperform those on the market.

The takeaway

If this all sounds complicated, that’s because it is. But the open and API-based world we live in means that businesses have the option of working with specialists, without having to restructure their tech stack.

The AI augmentation is capable of connecting to the existing infrastructure via API and acts as a layer on top—flexible and gaining insights from various use cases. Businesses and partners can identify AI implementations on a case-by-case basis, onboard quickly and see immediate results. No training is required, which allows businesses to [adapt] quickly.

The rapid time-to-value is driving AI adoption. The timeframe for converting a call (or text input from emails and chats) into structured data and into a signal that can inform a decision or action is a lot tighter than businesses are used to from other tech implementations and upgrades.  

Not all AI is created equal, and not all AI specialists have the same philosophy, expertise or awareness. Deciding which specialist to use is an important part of the AI journey.