"AI is the new electricity. Just as electricity transformed almost everything 100 years ago, today I have a hard time thinking of an industry that I don't think AI will transform in the next several years."
This already feels like old news by now – we can all see AI’s potential impact but what is not clear is how to actually start using it today – for this exercise ChatGPT doesn’t count.
AI removes constraints around many types of human-based input and output (ie. analysis, communication, cognitive recognition, creativity). So it follows that if we get it right, we could remove human-based constraints around these activities and increase output significantly.
But there’s a catch. Introducing any new technology into an existing operationally-complex company will have an impact on operations and require consideration before acting. AI’s wide-ranging capabilities and the emotionally-charged rhetoric surrounding it create additional challenges. Companies may be faced with fearful employees who feel threatened by it; general confusion over what is really possible; and how to actually “do AI.”
Impactful? Yes, but also disruptive.
How can companies take action?
On this subject, Ng has offered an “AI Playbook” – more like a framework for company leaders to start working with AI and applying it to their business. Overall, he stresses that a positive first experience is more important than maxing out the value created to gain internal momentum.
Identify a project with a high chance of success - This is more important than picking the highest value project.
Demonstrate traction within 6-12 months
Use internal or external technical resources but be sure to pair them with business expertise to insure solutions address real problems
Ng suggests undertaking a minor AI project before trying to craft a comprehensive AI strategy. He believes that the hands-on experience and familiarity with AI capabilities is crucial and significantly increases the chances of long-term success. This aligns with my personal experience.
At Jukin Media, we approached our first AI/ML project in a similar manner but we were unaware of the AI Playbook so I guess it was just luck.
One of our operational challenges was creating metadata for the videos the company acquired for its library. Since the company’s inception it was done “by hand” – the content operations team would watch each video as they came in and create the tags by hand in the internal DAM. The tags would be a mix of emotional/sentiment and hard descriptors to enable search and distribution functionality for the internal teams and partners.
Unfortunately, the process would bottleneck when the team fell behind (volume spikes, vacation days etc) and there were times when the team would need to “re-tag” videos for some attribute that wasn’t deemed important at the time of ingestion. As the library grew, this became a costly activity.
We knew that we couldn’t automate all metadata creation but if we were able to reduce the human workload by more than 50%, we would consider it a success. At the time, computer visioning technology wasn’t capable of understanding sentiment or emotional context which, for Jukin, was an important characteristic.
We focused on automating object detection and leveraged Amazon’s Media2Cloud APIs to build an interface that could create tags (and timestamps). We used an outside development partner with experience integrating the APIs and to build the front-end. In 3+ months, we had a working prototype and began to get “hands on” with code and its capabilities which led to a bunch of ideation and future projects.
The key takeaway was focusing on task automation of an existing workflow – not trying to replace a role or department. At this stage of the technology, it’s a more realistic approach.
To think through opportunities in your company, I suggest the following approach.
>>Get Familiar with AI Capabilities
The current state of AI and Gen AI excels at the following (and more);
- Text – Analysis and summarization
- Visual Recognition – pattern matching
- Creative Expression – visual and text generation
- Data Analysis – pattern recognition
There are free resources (video, podcasts, articles) everywhere to help you on-board. I found this very useful but a few Google searches will turn up a lot of great resources.
>>Review Internal Operations
Once you get a sense of the existing capabilities, review your internal workflows and try to identify tasks or work deliverables that would benefit from AI automation. Start small. Do not set out to boil the ocean and automate entire jobs or departments – it’s highly unlikely to be successful as a first implementation. Include your business experts and pair them with your technologists (internal or external) to ensure creating business value is at the forefront of the conversation.
>>Revenue Opportunities
On the revenue side, review your product or service offering and how it serves the needs of your customers. Does AI unlock opportunities for adjacent offerings that would solve additional needs? The answer is yes.
Admittedly, this is a more involved process that has its own framework (my favorite is Jobs-to-be-done aka JTBD) but whatever you’re currently employing can work. AI capabilities can (and will) create new product and service opportunities. If you don’t do it, someone else will.
To learn more about to get started with AI in your business or any other topic I’ve written about, connect with me via LinkedIn or set up a call.