[Interview] Ben Sigerson, Converseon: Solving Data Quality Issues with AI
On the 18th of June, we’re showcasing the newest, biggest, and best social intelligence tools, and insider tips on selecting the right social data tech at our first Social Intelligence Tech Demo Day.
You’ll find Converseon there demo’ing their AI-powered text analytics tech. Ahead of their demo, Ben Sigerson shares some insight into what to expect, why AI-powered text analytics can help you to increase the complexity of the questions you answer with social data.
Who are you and what are you doing at demo day?
I’m Ben Sigerson, the VP of Solutions at Converseon and I’m here to introduce the advantages of using AI / machine learning models for social listening. The impact of this AI-powered approach is far-reaching, from increased sentiment accuracy to more effective reporting on complex brand health and CX measurements, where there’s no room for data quality issues or manual data cleaning.
Who should come to check you out?
People who are involved in the measurement or analysis of a brand's performance. Typically we speak to social listening professionals, data scientists, marketing executives, corporate communications--anyone interesting in using AI / Machine Learning to simplify complex research.
Why is now the time to be investing in social intelligence (and your solution!)?
Social data can present a real challenge in getting a proper read of how a brand’s reputation and performance is being perceived, but it is a crucial source of consumer opinion. We firmly believe that using AI assisted, machine learning models is an effective way to tackle this problem at scale. These models understand the nuances and changing lexicon of conversation and can be combined together to provide measurements for different brand attributes, ultimately rolling up into a Brand Reputation score - a much more reliable and palatable way of communicating performance with leadership and different parts of an organization.
During the recent pandemic, we saw that the American Association of Advertising Agencies ran a survey, concluding that 40% of consumers want to hear how brands are responding to the outbreak; a large proportion of consumers want an organization to take a stand and have a point of view. Typically we see the volume of social conversation and sentiment being used as a crude measure of brand performance, but this provides very little useful insight and is often misleading as there are just so many false positives hidden beneath the surface.
What one word sums up your business?
Efficiency. If you consider all of the solutions Converseon provides -- AI-powered sentiment & emotion analysis, machine learning models (“classifiers”) for measuring brand health attributes/topics, the voice of the customer, and spotting consumer trends, and custom BI environments for quickly making sense of the output from these AI models -- the value of all of it boils down to efficiency.
Social media, news media, and other text data tends to be unstructured and messy, making it difficult for traditional research methods to extract reliable insight. Bringing AI to bear on this problem, as Converseon does, massively simplifies this process for all, from the analysts working with data directly to the executives who need actionable insights fast.
For example, one enterprise client used our AI models to reduce their time spent on weekly reporting by 80%, while also increasing the quality of the insights they were able to gather. By measurably demonstrating value, they were then able to win the trust of other teams in their organization. In a short period of time, usership of Converseon’s AI-powered approach expanded from just the corporate social listening team to all five lines of business.
What’s the best use case(s) for your technology?
We specialize not only in applying AI to research, but also in building intricate measurement frameworks, so that all of our AI models work together cleanly. Some use cases include:
- Measuring and benchmarking competitive brand health on key attributes
- Understanding drivers of positive and negative customer experience
- Trend-spotting: identifying emergent topics you may not have known to measure, and linking them to emerging, industry-specific trends
In complex use cases like these, you need multiple systems of measurement to work cleanly together--sentiment, topic identification, attribute measurement, identifying the voice of the customer accurately, etc. This is where traditional methods often fall short: keyword-based social listening and rule-based sentiment analysis often cannot deliver the accuracy you need, and survey-based approaches are often too slow, forcing you to take a rear-view approach that leaves you out of touch with current trends.
What makes your technology and/or approach unique? What’s your USP?
Converseon relies exclusively on AI / machine learning approaches for extracting insight and value from social media and other voice-of-customer data. This alone makes us a distinctive, niche player in social listening and market/customer insights. In social listening, for instance, the keyword is still king, insofar as it is still the main tool that social listening professionals rely on to segment social data.
In the last few years we’ve seen some vendors start to provide AI-powered approaches, and brands are increasingly hiring their own teams of data scientists to build custom AI in-house. But many of these recent adopters are finding there’s a difference between simply adopting AI and implementing a practical vision for AI that delivers results.
Converseon has been focused on social data since the mid-2000s, and on AI/machine learning specifically since 2011 or 2012. I think this has given us a big head start in terms of honing our approach; in the last 8-10 years we’ve gained a sophisticated understanding of how to bridge that gap between “having AI” and practically implementing AI in a way that delivers real business value.
What's the best thing about working in social intelligence?
I think there’s a widespread understanding or feeling that despite all of the growing sophistication of social listening approaches today, this dataset’s potential to provide insight still remains largely untapped.
It’s been accepted for some time that social listening data can be used to make predictions in lots of different areas; politics, consumer behaviour, financial and brand performance. This is what our clients and partners really want and where social really becomes intelligent. Getting the data in shape to apply predictive analysis can take time, even if you make use of “bleeding-edge” AI and machine learning models. But there are so many bright minds finding faster and better ways to get there. “Speed to insight” has become a race--the person who gets to that predictive insight first in order to take an informed action, is likely to be the winner.
That’s an exciting place to be, because there’s still a lot of innovation and change happening, and the “AI-for-Social-Listening” market is not at all commoditized yet.
Learn more about SI Tech Demo Day, and access all the presentations from this year's event, on demand