Skip to main content

Making sense of generative AI for your team, your company, and your future

Supercharge your customer journey with custom AI chatbots!

Tech executives seem to only appear before Congress when compelled by invitation or pressing current events. But this spring Sam Altman, the CEO of OpenAI, visited lawmakers proactively. He testified about the need to regulate artificial intelligence (AI), which his company’s explosive ChatGPT product kicked off.

At Sendbird, we’ve embraced conversational generative AI to help businesses engage, support, and monetize online effectively, even without human presence. Using OpenAI’s ChatGPT and Google’s PaLM2, our engineers create more engaging and personalized chatbot experiences — just one use case.

In this piece, we will cover the context of generative AI and share some best practices for building its capabilities into your business — safely, efficiently, and proactively.

Not sure where to start with generative AI? Here is our guide to the basics

Sendbird’s perspective of AI is that your businesses should futureproof for the inevitable. Let’s start with the basics:

What is generative AI?

Generative AI is a specific type of artificial intelligence that creates original content using inputs from existing content. IBM’s research blog describes generative AI this way:

“Generative AI refers to deep-learning models that can take raw data — say, all of Wikipedia or the collected works of Rembrandt — and “learn” to generate statistically probably outputs when prompted.”

Generative AI refers to a branch of Artificial Intelligence that involves creating models capable of generating new content, such as images, text, or audio, that closely resemble examples from a given dataset. Generative AI models use techniques like deep learning and neural networks to generate original and realistic outputs.

Who are the key players?

As of June 2023, three major companies dominate the generative AI space: Microsoft, Google, and OpenAI. In 2019, Microsoft invested $1 billion into OpenAI and doubled down with a massive $10 billion investment in early 2023. 

Hundreds of startups are cropping up in the space as well. These companies are packaging generative AI technology into consumer-facing technology designed for specific target audiences. To help people keep up, Dealroom maintains a list of generative AI startups.

What are the key AI products?

This is a fast-moving space, and new products are always in development, but these are likely the ones you’ve already heard about.

OpenAI released ChatGPT in November 2022. This AI chatbot is built on OpenAI’s GPT-3.5 and GPT-4 foundational large language models.

In response, Google released Bard, an AI chatbot built on the Pathways Language Model 2 (PaLM 2). (Initially, Bard was constructed on the LaMDA family of large language models.)

OpenAI also released DALL-E and DALL-E 2, which generate digital imagery based on prompts or “natural language descriptions.” Midjourney offers a similar product. The ethical concerns around fake imagery are obvious; a more light-hearted example is a deep fake of Pope Francis wearing a white puffer coat.

What are top generative AI business applications?

As generative AI develops, new use cases will pop up every week, or faster. Instead of keeping an exhaustive list, we recommend developing a framework to help sort through the news, pitches, and marketing — so that you can find the applications that will matter most for your business.

As an example, consider KPMG’s April 2023 report on generative AI models (PDF, 345 KB), which places generative AI applications into five categories:

  1. Content generators: Create blog posts, emails, social media posts, ads, and more

  2. Information extractors: Turn long-form content, whether legal documents or news articles, into short, easy-to-consume summaries

  3. Smart chatbots: Hold personalized conversations with customers

  4. Language translators: Translate languages accurately and quickly

  5. Code generators: Turn natural language inputs into pieces of code or applications

The background of generative AI

Another way of conceptualizing generative AI in business is to examine the technology’s background. How have businesses communicated with customers over the decades, and how has the overall marketplace changed? Generative AI is built upon these past developments — and understanding this legacy can help you stay agile in the present.

18th century

In this century, local businesses were built through personal interactions. Because mass advertising was virtually impossible, word-of-mouth marketing was critical for survival. Face-to-face conversations between merchants and customers helped as well, and a business could also create printed signs and “notices” to advertise goods and services.

19th century

Congress passed the Postal Service Act in 1792, which established the Post Office Department (which became the United States Postal Service in 1970). This passage meant businesses could reach even more customers, and a print media industry of newspapers and magazines developed throughout this century. The newly established postal service also allowed for direct mail, including catalogs, brochures, and promotional letters. For example, the first of the classic Sears, Roebuck, and Co. catalogs circulated in 1894.

Early 20th century

Alexander Graham Bell invented the telephone in 1876, and in the 1900s, businesses started using the phone to take orders and provide customer service. As radio broadcasts became popular, businesses took advantage of the opportunity to enter consumers’ homes and daily routines. Company owners also used out-of-home advertising, including billboards, posters, and signs in public spaces.

Middle 20th century

Direct mail and radio continued, but by the mid-1900s, businesses had a powerful new medium to try: television. The first TV ad appeared on July 1, 1941; the 10-second commercial advertised Bulova watches before a baseball game between the Brooklyn Dodgers and the Philadelphia Phillies. As the platform developed, companies eventually started running product placements in popular TV shows. Telemarketing also developed during this time, and “do not call” protections weren’t established until the 1990s.

Late 20th century

The programmer Ray Tomlinson sent the first email using the “@” symbol in 1971 — setting the stage for email marketing to develop in the late 20th century. Businesses could easily and inexpensively promote products to customers worldwide (the fax machine helped, too). Customers gained easier access to businesses as well with toll-free numbers for placing orders and getting support.

Early 21st century

The final turn in this story, of course, is the internet. Websites allow businesses to build their brands and promote goods and services. Online forms and, later, chatbots enable them to communicate with customers. Social media developed in this era; customers could now connect with businesses faster, and businesses could answer questions, share updates, and advertise on these platforms. Many businesses also invested in mobile apps. These allowed for more personalized, branded customer experiences than SMS marketing, which also developed during this time.

The case: Generative AI’s potential for good

Earlier this year, the YouTube channel Economics Explained published a video essay with the (inflammatory) title Did Washing Machines Change the Global Economy More Than the Internet? The creators noted that washing machines and other technologies automated domestic household tasks. This freed up many women to enter the workforce and add to the economy — a shakeup, but one that created a boom.

The positive potential of generative and other forms of AI is that something similar could happen: AI tools could automate mandatory but repetitive tasks for professionals across industries. Suddenly these workers have free time, energy, and capacity to create. Who knows what they could do?

Suchi Srinivasan of Boston Consulting Group calls generative AI a great equalizer. In her words, enterprises don’t have to have “armies of data scientists” to capture value from artificial intelligence. KPMG’s report elaborates on this value capture, noting that the large multimodal models (LMMs) powering applications like ChatGPT could eventually:

  • Summarize and classify legal documents

  • Respond to customer questions

  • Support business advisors

  • Create drawings for engineers and architects

  • Build business reports and pitches

  • Develop code for software applications

A Salesforce 360 generative AI blog post imagines several more generative AI use cases specifically for sales teams: virtual assistants for sales representatives, replacing lead forms with personalized engagements and automating the prospecting process.

The counterpoint: Generative AI’s potential for harm

Jokes about AIs taking over the world will likely never end. But, there are serious ethical issues with generative AI — part of why OpenAI’s CEO visited the Hill, and why 355 industry executives signed the aforementioned open letter.

Boston Consulting Group’s (BCG) article on generative AI elaborates on a few of these risks:

  • Unknown and/or unexpected capabilities: New uses for AI are being developed every day apart from developers’ involvement. This means that, without “the right guardrails,” generative AI could be used in ways the experts who built it never imagined and can’t govern.

  • Bias and toxicity: Many of the most popular language models (which power generative AI programs) are trained on the internet, which is full of toxic and biased language. This means the AI outputs could be biased, too.

  • Data leakage: The information put into programs such as ChatGPT isn’t necessarily private; it could get re-incorporated into the system and become public knowledge — a problem for any sensitive information shared.

  • Hallucination: The misinformation threat is huge with generative AI, especially programs with text output. Many have documented that ChatGPT sometimes makes convincing arguments about completely wrong positions.

  • Lack of transparency: Developers typically do not disclose the data that AI models were trained on — which makes it challenging to fact-check ChatGPT, for example.

  • Copyright controversies: The most popular AI models today are trained on text from the internet, which was, of course, written by millions of other individuals. BCG posits a legal question: Does the content created by these AIs “amount to duplications of copyrighted works?”

Best practices for leveraging generative AI responsibly in your organization

You may feel pressured to start incorporating generative AI into your processes quickly. Many of your teams may already be using these tools. We believe the key to using generative AI responsibly and ethically is to start with a companywide framework, one guided by multiple, expert voices within your company. Drawing on the sources referenced above, here are four starting points for responsibly leveraging generative AI:

  1. Thoughtfully choose your business use cases, instead of building them in at random.

  2. Pick a platform that prioritizes security and privacy.

  3. Keep an eye on relevant, new, enterprise-grade offerings and have a plan for evaluating them.

  4. Establish an enterprise-wide framework to guide the development of use cases at your company.

Also, check out these resources on responsible generative AI:

Ebook Grow background mobile

Take customer relationships to the next level.

Ready for the next level?