Discover more from Sparrow One
How generative AI could impact knowledge work
Companies need a POV and a plan
Until recently, when someone mentioned artificial intelligence (AI), that may have involved offshore contract workers mimicking AI-like functions rather cheaply. AI has been overhyped for so long, but the promise and reality have always been very far apart—until now. With advances in large language models and new tools like ChatGPT, that promise and reality are getting closer. There's been a legitimate technological evolution, and that's why the technologists are excited.
All the recent headlines about ChatGPT—It’s sentient! It’s belligerent! It could destroy the world!—have made AI (and Bing) suddenly all the rage. There are also some dark undertones to the news coverage suggesting that AI is going to replace humans. But the reality is that when this type of technology is layered into most knowledge work, its first applications will act in an assistive way, enabling professionals to do their work more effectively. And then come the fascinating second-order effects. For example, when technology is democratized for smaller companies, that mom-and-pop shop that previously didn’t have marketing as a function can now have marketing as a function.
Thanks for reading Sparrow One! Subscribe for free to receive new posts and support my work.
Yes, it’s automated marketing, but that's still a step forward, especially in the platform era. Companies like Google and Facebook have made it easy for lay people to advertise online and leverage their platforms’ business functions. Generative AI has the potential to make a similar impact in marketing and advertising. Since its algorithms can generate content—including text, images, or videos—in response to queries or prompts, generative AI presents many possibilities for marketers and advertisers that at this point can be classified into three general and immediate buckets.
Creative development: Can you use generative AI to improve your creative and production process? We can easily see a tool like Midjourney enhance production for those tasked with pumping out lots of creative, perhaps helping them quickly evaluate their options without the need to shoot entire scripts. Instead, AI could mimic the finished product so that humans can get quick feedback from clients or consumers and iterate faster. AI could also make possible entirely new types of creative or capabilities that were once cost-prohibitive.
If you're a brand marketer or creative agency, it’s hard not to be excited about the possibilities. At the same time, AI will not fix everything, and in some cases, the results will be likely underwhelming. Those who work on the creative side have to balance testing AI technology to get good enough for it to pay off with incorporating this type of learn-as-you-go process into their regular workflows.
Assistive capabilities: As we mentioned, new technologies often begin in an assistive fashion. Could generative AI also fit into client-facing communications, such as AI-assisted email correspondence? This function has been around for a while in sales assistance tools and lead-gen tracking, but maybe AI can now be applied to ALL professional communications. One reason why everybody's so excited about current-gen team collaboration tools like Slack is because they allow you to create triggers and easily programmable escalation paths. For example, if a client takes a desired (or adverse) action, you can create an alert that informs specific people immediately so that the company can continue moving the client along the customer journey. Although we've been using automated assistive technologies for a while, generative AI may represent its next evolution.
One of our favorite examples of an assistive AI tool is Otter.ai. We’re received so much value personally from this AI-driven speech-to-text transcription application. Otter makes it very easy to record conversations, meetings, or interviews and create near-instant notes and summaries that you can then collaboratively edit. The transcriptions aren’t 100 percent accurate, but they’re pretty good and save us a lot of time.
Work augmentation: Perhaps the most exciting aspect of generative AI is its enabling potential for the small business owners driving the economy—and the bottom lines of larger tech platforms. There are roughly 32.5 million small businesses in the US, representing 99.9 percent of all businesses in the country, according to the US Chamber of Commerce. Generative AI can provide new capabilities for smaller businesses that many couldn’t previously access.
Copy.ai, for example, can help small businesses write marketing copy. If they wanted to write an email newsletter to their existing customer base, the technology could package and productize it in such a way that a lay person could easily run a spring discount campaign to attract new customers or launch a referral email to existing customers to drum up new business. Once the campaign runs, the tool provides results that the user can analyze to inform future campaigns. Copy.ai is actually a twofer—both an assistive technology and primed for small businesses.
Developing a new set of skills
We're now at the point where large language models and similar technological advances can do most of the heavy lifting for some tools—but not quite 100 percent. That’s because we haven’t reached consensus on how to approach the underlying data needed to train AI to ensure an accurate output. There are many examples of ChatGPT getting it wrong. When asked about the largest country in Central America that isn’t Mexico, for instance, it incorrectly responded with Guatemala, not Honduras. The chatbot has also been known to make up things to fulfill queries, but it might be hard to tell because the answers are grammatically accurate.
What this means is that marketers, advertisers, and other knowledge workers will need to develop a new set of skills around creating good prompts so they can get back the kinds of results that they can actually use. This isn't dissimilar from being a competent Googler; there's so much garbage on the internet that if you don't have a high degree of media literacy and know how to craft and filter what you're searching for, you'll likely wade through SEO-gamed junk instead of finding the results you need. It’s the same phenomenon with AI prompts. Depending on how you craft your queries, you may get very different results. There is a learning curve that people need to embrace and find the time to experiment.
Marketing and advertising professionals shouldn't be sitting on the sidelines. We need to roll up our sleeves and experiment with these technologies to figure out how to include them in daily workflows. The good news is that it's easy to start experimenting with AI and ML. It’s relatively inexpensive to start—think tens of dollars per month. Practice writing prompts.
More than likely, the tools marketers already use on a daily basis will have some sort of AI offering. Adobe just announced generative AI tools through Firefly, and Google is opening up its chatbot Bard for early access. Create a plan to test AI tools with the expectation that the initial results will likely be underwhelming. It will take time to train users to get better—and for the AI tools to improve and get better acquainted with what humans need.
There’s really no alternative to putting in the work and getting smarter on these topics fast. Marketers and advertisers shouldn’t take vendor announcements at face value. They need to educate themselves and develop an internal understanding of AI and point of view—or find someone to help them come up with one. What does the leadership team need to know about generative AI? How should they prioritize it, and what do they act upon?
With Google and Microsoft racing to bring AI to market, does generative AI put search back into play? ChatGPT certainly has a lot of people excited about Bing, which is quite a feat. Googling is now a verb, Bing far from it. With AI-driven search engines, companies trying to claw their way up the Google search page will get disrupted. And with AI-driven search engines, we could also be at the cusp of breaking reality completely. What that means is that AI-driven search engines may eventually create a custom set of results for us that would be very different from the results for you. Basic facts such as the number of planets in the solar system typically have a finite set of responses, but if algorithms are pulling from different sets of potentially questionable training data to answer queries with little to no transparency, we risk losing whatever little consensus we have of what's true. This brings us to one of the biggest challenges we must overcome with this evolution of AI: training data.
There has been a lot written about the ethics of AI algorithms, but the underlying training data for AI has immediate practical implications for many companies. Where is the training data sourced from, and is your company's data or intellectual property being used as input into these generative algorithms? If so, do you need to do something about it?
Getty Images recently sued Stability AI, the startup behind Stable Diffusion, an open-source AI art generator, for allegedly copying more than 12 million Getty images that it used as training data to create a competing business without permission or compensation. Is it fair for an AI company to use another company’s IP to train their algorithm and subsequently commercially benefit from the output? That possibility requires companies to revisit their commercial and IP rights. We saw something similar with the rise of streaming when film and television writers found themselves in the unenviable position of creating content that was monetizable in non-digital channels, but all of a sudden in digital there were no guardrails or commercial mandates that would value the work and pay the creators accordingly.
The AI roadmap
AI should be on everyone’s radar, in every role and at every level. Companies need to identify the issues they should consider and the steps needed to protect themselves as AI evolves. Companies should have someone in-house to own AI. In larger companies, it will probably be several people across different departments working on AI on their own before coming together to advocate for a more formal policy. The AI point person or team should also consider how to ensure that the company is at the forefront of testing AI and understanding how to operationalize it. A great first step would be a crash course in what your company’s executive team needs to know.
When vetting generative AI tools, companies should ask about the source of the training data. That will offer clues about the appropriateness of commercially using the tools and provide signals about any potential biases that you may need to counterprogram for. Knowing the input could also provide a high-level view of how the algorithm achieves its output and what it takes into consideration.
You’ve likely seen ample debate about the potential for AI to take over, but should that even be a goal? Tesla’s Autopilot feature inspires some to envision no longer driving themselves anywhere, instead getting chauffeured everywhere but not having to pay a (human) chauffeur. But is that what we really want? We would argue that for most people, the goal is a bit different. Maybe they just want their repetitive and predictable daily commute to be more automated—rather than every trip they take, which is a different and much more challenging problem to solve.
In other words, some of us are getting ahead of ourselves. Let’s take a step back and think about the more tenable and practical problems that AI could solve for many people today—not 10 years from now. As professionals, we shouldn't be scared about the prospects. Sure, we understand how it might be a little unsettling for, say, a journalist to read about CNET’s plans to use an algorithm to write some of its stories. But what if it’s automating the creation of stories that no one has the bandwidth to cover? (CNET’s plan ended up not going so well.) While some of the recent hype around AI borders on sci-fi, the day-to-day reality of generative AI is much different.
It's more about getting a little smarter or quicker for your repetitive tasks. Rather than automating 100% of your busywork, how about 35% of it—and gaining two or three hours back every week? That's a huge win for many knowledge workers. What does that kind of efficiency free us up for?
How would you use the free time you gained by adopting AI tools in your business? How much is your time worth?
Thanks for reading,
Ana, Maja, and the Sparrow team
Enjoyed this piece? Share it, like it, and send us comments (you can reply to this email).
We’re a results oriented management consultancy bringing deep operational expertise to solve strategic and tactical objectives of companies in and around the ad tech and mar tech space.
Our unique perspective rooted deeply in AdTech, MarTech, SaaS, media, entertainment, commerce, software, technology, and services allows us to accelerate your business from strategy to day-to-day execution.
Founded in 2015 by Ana and Maja Milicevic, principals & industry veterans who combined their product, strategy, sales, marketing, and company scaling chops and built the type of consultancy they wish existed when they were in operational roles at industry-leading adtech, martech, and software companies. Now a global team, Sparrow Advisers help solve the most pressing commercial challenges and connect all the necessary dots across people, process, and technology to simplify paths to revenue from strategic vision down to execution. We believe that expertise with fast-changing, emerging technologies at the crossroads of media, technology, creativity, innovation, and commerce are a differentiator and that every company should have access to wise Sherpas who’ve solved complex cross-sectional problems before. Contact us here.
Thanks for reading Sparrow One! Subscribe for free to receive new posts and support my work.