Helping Small Teams have a Big Impact: ActionVFX Customer Story
As a leading provider of visual effects footage (think explosions, gunshots and tidal waves) managing a huge catalog of visual products, the team at ActionVFX value nimble and efficient workflows. For Chief Operating Officer Luke Thompson, harnessing the power of AI was the game-changer he was looking for.
From automating tedious content creation tasks to streamlining complex business operations, Relevance AI has empowered the ActionVFX team to scale their operations with a small team. Here, we'll dive into how Luke leveraged Relevance to build a suite of custom tools that have transformed the way he works.
But first, what is an AI Tool?
AI tools are how you build custom integrations, LLM prompt chains or traditional automations in Relevance. You can give these tools to your agents, or use them yourself via an auto-generated app. When building AI tools in our platform, you can easily use LLMs, make API calls, access our pre-built integrations or even run custom code.
Optimizing SEO with a a suite of AI tools
Following an SEO audit of the ActionVFX site, Luke zeroed in on several content generation issues that could be impacting search engine visibility and rankings.
Top among these were:
- Non-descriptive H1 titles and meta descriptions.
- Missing resources such as product FAQs.
- Thin product description text.
ActionVFX faced a key challenge creating SEO optimised content at scale - the platform their website used didn't automatically generate essential elements like meta titles and H1 tags.
As a first step towards improving the discoverability of their extensive product catalog, Luke built a tool to optimize H1 titles by updating them to reflect on page content.
Luke established a solid foundation of structured data by using the ActionVFX sitemap to pull hundreds of URLs for product pages, blogs, and other resources, bringing these into Relevance as a data table. He then ran a bulk scraping process to capture the text from each page, and with this comprehensive dataset in hand, he was able to start building.
Luke used very straightforward prompt in the LLM step for his tool:
“The goal is simple: Take a scraped webpage summary and write a H1 meta title based on best practices for SEO and the contents of the page. Only reply with the meta title.”
By running his customized H1 title generation tool in bulk across the company's product data table, Luke effortlessly produced hundreds of optimized page titles. He was also able to apply this process to writing meta descriptions and product FAQs.
Rather than dedicating valuable resources to manual content creation, he used the speed and scalability of Relevance AI to deliver an impressive number of high-performing assets.
Generating Rich Product Description Text
The team at ActionVFX faced another challenge common for e-commerce businesses – their product pages were lacking the robust, informative content needed to engage customers and perform well in search engine rankings. With only basic product titles and sparse descriptions populating their webpages, the content was deemed too "thin" to represent value to a search engine.
To solve for this, Luke built a Product Description Generator tool that could automatically create expanded, in-depth content to replace bare-bones information.
The first step was to gather the necessary inputs – the product name, the existing product description, and any relevant keywords. Luke then crafted a detailed prompt for the selected LLM to work with, instructing it to use the provided product data as a foundation.
By running this prompt against their catalog of product information, the tool was able to generate expanded descriptions that breathed new life into the thin content - a massive improvement compared to the original product specs.
As LLMs often adopt a generic or unnatural tone in their writing, Luke built an additional step using Claude 2.1 to "de-ChatGPT" the descriptions, further refining them and smoothing out any robotic phrases (pun intended).
As a final step, Luke built in a fact-checking safeguard, which referenced the original product information to verify the accuracy and technical soundness of the generated descriptions. This provided an extra layer of quality assurance, solving for any potential hallucinations or inconsistencies.
The end result was a comprehensive, multi-step workflow that transformed ActionVFX's thin product content into detailed descriptions designed to appeal to customers and search engines.
The results *drumroll*
With the suite of AI tools he built, Luke was able to instantly fix a mind-blowing 4,614 on-page SEO issues, raising ActionVFX's overall on-page SEO score from 40 to an impressive 72. This improvement has paved the way for a significant boost in organic search visibility and traffic.
In short
By embracing the flexibility and customization of the Relevance platform, the ActionVFX team was able to enhance their content discoverability and deliver a superior customer experience. Their journey with Relevance serves as a powerful example for other companies looking to drive growth with AI tools, and as they now explore the integration of their custom tools with intelligent agents and AI agent teams, we’re excited to see them further scale their business with the power of an AI workforce.
