These days it seems like there’s an AI tool for everything. Now there’s even an AI tool to manage your AI tools.
On Thursday, artificial intelligence startup Anyword released what it calls a “gen-AI performance platform” that analyzes and scores all of the content generated by the various tools in a marketer’s stack.
“We think you’re going to be using multiple AI applications and platforms, not just one, and we want to help you do that,” CEO and Co-Founder Yaniv Makover told AdExchanger.
The platform has integrations for CRM systems and social ad channels, as well as integrations with AI models, such as ChatGPT. It’s able to advise marketers about what type of data each model will need in order to give the best results.
Tailored content
The company was launched in 2021 (just ahead of the generative AI boom, Makover noted) out of a machine-learning-focused company called Keywee.
Using Keywee’s suite of tools, media companies like The New York Times and NBC can analyze their news content to predict what audiences will respond to, but not to influence the writing itself. The technology wasn’t there yet at the time.
Anyword was built to fill that need by ingesting all of a marketer’s publicly available content – ad accounts, social media pages, websites, email campaigns, etc. – and then make suggestions about how to tailor new content to a specific audience.
The platform does this using retrieval augmented generation (RAG), which improves an AI model’s output without requiring elaborate prompts from the end user by retrieving additional information and context from a large database before generating content.
Anyword’s RAG was trained partially on an enormous test data set that the company already owned and partially from scraping publicly available performance data (although not any copyrighted content, Makover stressed).
Data from users of its free tier is used to improve the model, but not data from enterprise clients like Vimeo, Dollar Shave Club or AlphaSense, which are siloed off into their own private versions.
So far, the company claims to have driven a 20% increase in sales for B2B, B2C and enterprise marketing teams using the platform. Clients are also able to track different metrics depending on their larger business goals, such as cost per acquisition or click-through rate.
Art vs. science
But Anyword’s primary selling proposition is that its platform will be able to help marketers achieve a consistent, easy-to-control brand tone across all channels, particularly across various AI models.
Of course, a full-time writer can’t help but ask the question: If so many AI tools and platforms are needed to ensure a consistent brand tone, wouldn’t it be easier and cheaper to hire a writer with brand marketing experience instead?
Makover had an answer for that.
An AI model can’t outperform a professional marketer, he said. But what the tool can potentially do better than humans is back up its copy suggestions with quantitative data.
To test that hypothesis, in 2019 the Keywee team gave content that had already been tested for a particular target audience to a number of copywriters and asked them to predict which would perform the best.
Apparently, those copywriters could only predict the best-performing copy 60% of the time, often disagreed with one another and even contradicted themselves when shown the same copy three months later.
Despite this, Makover acknowledged that a marketer using Anyword’s platform might have a reason to select a lower-performing copy variant. The important thing is that marketers have the metrics to make informed choices rather than only relying on instinct.
Co-Founder Omer Rabin, whose appointment as Anyword’s chief revenue officer was also announced on Thursday, recalled a similar conversation he had with a brand client about Anyword’s tools.
“None of you came to be a marketer here to focus on the science,” a manager at the brand reportedly told their employees. “By bringing a solution like [this], we cover the science so you can focus on the art.”
Part of the workflow
AI’s overall role in the average worker’s day-to-day offers up a similar contradiction.
Although most tools are designed as time savers, a recent Upwork survey suggests that most employees feel the exact opposite – that AI tools have actually increased their workload.
When asked about this, Makover did not seem surprised.
“I think there’s a lot of pressure coming from executive teams to adopt AI, and that’s at the expense of using best-in-class tools,” he said, further noting that it’s difficult to sort between platforms dependent on generative AI to function, like Anyword, and those that are merely trying to capitalize on the AI hype.
Still, he remained optimistic that the learning curve for AI adoption will quickly even out. “The applications are getting better and more useful, and I think that’s going to continue,” Makover said.
