InferenceCloud Quick Start Guide

Thank you for joining us in the journey to build the next generation of AI for marketing and communications.

Issues, suggestions or other information can be provided at any time at [email protected].

For quick optimization of your results, these callout sections contain essential tips and best practices. Simply compare your input against each piece of advice here, and we'll help you enhance your output.

</aside>

This short guide outlines some basic steps for getting started with InferenceCloud by following a relatively simple workflow using a real-world example to build data-driven strategies and then create content that is tailored to achieve a communication objective with a specified audience.

Overview of the InferenceCloud workflow

Research, plan and create – step by step

Quick Start Guide IC Workflow.png

Step 1 - Identify some data-driven topics with Strategy Explorer

You don’t have to start here, but most projects start with a communication objective and an idea of who you want to talk to (the Audience).

Start by creating a new Strategy Explorer notebook to easily crunch data and identify some topics for your communication project. The system will walk you through it so that you can hone your objective to a certain audience and identify the Topics that resonate with that audience the most based on billions of data points (see note below on Topic selection).

These Topics are also needed later on when you want to create some content (or create a content plan/campaign) so that Inference Writer knows what to write about and how (and when to write about it if you’re also making a content plan in Strategy Builder).

After working through the process and letting the system crunch billions of points of data, you will end up with a visualization and tables that give useful insights into strategies and Topics to achieve your objective with a specified audience.

Untitled

</aside>

<aside> <img src="https://prod-files-secure.s3.us-west-2.amazonaws.com/b8c407ff-d582-4ee4-8190-b93a08ebca4e/332b3a27-6016-4d77-8e34-8f0fad0169dd/tip.png" alt="https://prod-files-secure.s3.us-west-2.amazonaws.com/b8c407ff-d582-4ee4-8190-b93a08ebca4e/332b3a27-6016-4d77-8e34-8f0fad0169dd/tip.png" width="40px" /> Selecting Topics and letting the data guide you during analysis: very often users only pick topics to deep dive into that they are already familiar with. However, we have found that the best discoveries and innovations come from letting the data guide you, rather than trying to make the data fit your existing assumptions. As such, it makes sense to not only select the topics that fit into your assumptions, but to also pick some topics that you might think are unrelated to your communication objective and audience. By doing so, you may find that the audience cares about some topics you ignored until now, providing you with deeper insight into your audience mindset. You may also find that some topics you assumed were important turn out to be not so important to an audience or achieving an objective. ****

</aside>

Step 2 - Create talking points in Insight Miner

Insight Miner offers the possibility to ground all of your content generation and content plans in trusted source materials. It’s not absolutely required to create content in Inference Writer or Strategy Builder, but we highly suggest using this functionality to increase consistency in your AI-generated content, ensure uniqueness, improve the quality and trustworthiness of that content and eliminate the risk for AI hallucinations in the content.

Start by creating a new Insight Miner notebook and select some source materials, which can be files (like PDFs, PPTs, DOCs, MP4s, and such) or URLs. These source materials should reflect the attitudes, opinions, facts, and quotes you would like your content to contain.

After Insight Miner processes the source materials, it will generate Talking Points which will be needed later to create content and build content plans (also see the info box below).

<aside> <img src="https://prod-files-secure.s3.us-west-2.amazonaws.com/b8c407ff-d582-4ee4-8190-b93a08ebca4e/332b3a27-6016-4d77-8e34-8f0fad0169dd/tip.png" alt="https://prod-files-secure.s3.us-west-2.amazonaws.com/b8c407ff-d582-4ee4-8190-b93a08ebca4e/332b3a27-6016-4d77-8e34-8f0fad0169dd/tip.png" width="40px" /> Best practices for URLs as source materials: If you do use URLs as source materials for your Insight Miner notebook, some websites actively block crawling and the result is that either the website can’t be processed by our system, or the output is meaningless. If this happens, the best practice is to download the webpage as a PDF and then upload that PDF into Insight Miner. Also, please note that Insight Miner only processes one webpage at a time, not entire websites, so if you enter the top-level domain, e.g., www.inferencecloud.ai, it will only process the content on that home page, not every page in that web domain.

</aside>

<aside> <img src="https://prod-files-secure.s3.us-west-2.amazonaws.com/b8c407ff-d582-4ee4-8190-b93a08ebca4e/332b3a27-6016-4d77-8e34-8f0fad0169dd/tip.png" alt="https://prod-files-secure.s3.us-west-2.amazonaws.com/b8c407ff-d582-4ee4-8190-b93a08ebca4e/332b3a27-6016-4d77-8e34-8f0fad0169dd/tip.png" width="40px" /> Talking Points are all of the important points that the AI has found in each source material you have processed. Sometimes these are paraphrased sentences, sometimes they are word-for-word extractions, sometimes they are quotes from individuals, and so on. The AI does not pre-suppose what the user who uploaded the source material would like to create, so it extracts everything meaningful that it can. Sometimes this results in Talking Points that may seem irrelevant to you, such as privacy policies from webpages, but rest assured that when you ask InferenceCloud to do something with these Talking Points, it will only use the ones that are relevant to your request.

</aside>