👬AI Twins

Your direct access to the consumers

Introduction

consumr.ai was built with an intention to empower users to gather more knowledge about what the consumers in their market need. So, we built reports around various consumer touch points throughout their journey from having a mindset to buying the product and then making it actionable.

However, we also realized that creating a defined structure for a report at a point in time can only tackle finite use cases. It is a known fact that businesses from various industries have to encounter various types of scenarios and we wanted to build a solution that can tackle most if not all. There are two ways that could be done: Connecting more sources, and building dynamic, non-rigid enablers that can allow the users to twist the data consumr.ai has access to.

While we have been adding newer platforms like Snapchat, TikTok, DV360, and other DSPs, to cater to the first problem at hand, Conversations was our answer to building enablers that can interpret the data with the most flexibility we can offer. Conversation with consumr.ai is really different from your regular LLMs like Claude, ChatGPT and Co-Pilot. It focuses on data intelligence drawn from the data sources of consumr.ai and then gives flexibility to the user to answer the questions related to their business problems. However, there are three core issues with it:

1. How will it figure out what source of information is critical to answer the question at hand?

2. How do I identify the answer and then convert it into an action?

3. How do I give the user the capability to converse with their consumers directly?

The answer may be simple to arrive at, but building it requires a lot of research and development:

Answer 1: We built Co-Pilot to tackle this issue. You can learn more about it in our upcoming article talking about Co-Pilot/Wingman.

Answer 2: This is a work in progress, and we will soon come up with an answer.

Answer 3: AI Twin is the feature developed just to answer this question for you.

In this article, we will talk about everything AI Twin.

What is an AI Twin?

An AI Twin represents your consumer. Consumr.ai can uniquely generate insightful report assets by analyzing recent behaviors, intentions, and conversations of your 1st party, 3rd party, or categorical audiences. Based on these insights, consumr.ai creates personalities that mirror a segment of your consumers, depending on their stage in the consumer journey. Brands and agencies can interact with these AI Twins, asking questions and receiving immediate responses based on almost real-time data derived from the insights. This groundbreaking technology has multiple applications, which we have detailed further in the article.

The Philosophy

One of the main objectives of consumr.ai is to make complex datasets comprehensible. The initial output, generated using data signals from Meta, was a one-page analytics report composed of multiple cards, each presenting a unique story to the user. Despite our efforts, we found that most users, while able to relate to the content, struggled to interpret and derive meaningful insights from it. To address this issue, we introduced conversational guidance. However, users were unsure of what questions to ask and wished to generate new reports from these conversations. This necessity led to the development of consumr.ai's CoPilot—a sophisticated model that trains itself on intricate datasets, identifies gaps, and autonomously generates reports to fill those gaps. Our efforts were focused on making extensive and monotonous data as digestible as possible for human users.

This endeavor gave rise to the concept of the AI Twin. We envisioned an AI that could directly understand consumer thoughts by embodying the personality of the brand’s customer and retaining the memory of real users based on their search behavior and online sharing. By integrating these two aspects, we brought to life an AI Twin that amazed us with its responses and the precision of its outputs, offering an almost conversational experience.

Why AI Twin?

The biggest issue with brands is that people who get to interact with majority of the customers are not generally who make the decisions on how the product or service should change. Something that has a direct impact on customer’s experience. Brands that go the extra way, ensure that they get in touch with a very small sample of the consumers that represents the masses. Again! Not the best option. But you can’t blame them. It is too time consuming and costly to round up thousands of consumers and ask them questions, something that surveys and focus groups does. Sadly, the consumer landscape in today’s time moves way faster than the speed at which these initiatives generate results. Bottomline… There is no solution.

AI Twin was built as a technology to enable decision makers in the corporate offices to talk to real customers anytime of the day or night, and any number of times. If you don’t have too many questions to ask… No problem! Visit the AI Twin anytime and continue where you left off. The beauty of this system is that life your regular consumer, the AI Twin knowledge auto updates every month. It’s not your average LLM that updates its knowledge base once a year. So, even if you converse with it every month, you will be conversing with a newer version of the twin. We will discuss more on this in the sections to come.

Benefits of using an AI Twin

There are a lot of benefits of working with AI Twin. Let us talk about some of them here. While you are working with your own AI Twins created on consumr.ai, please don’t hesitate to let us know of the new ways you discover of working with it.

1. Changing consumer landscape

In the past, humans took a century to realize that gunpowder could be used to make bullets. Its application was limited to making firecrackers. If you rewind time further, the application of fire was limited to using it as a source of light. It took us more than 100,000 years to learn that it could be used to cook food. And now, a vaccine that historically took seven years to create, was created within seven months. This shows the rate of speed of change, owing to hundreds of years of education, knowledge and millions of years of evolution. Owing to the technological advancement, humans have adapted to consuming GBs of information in a day through various digital sources and the technology of today does an extremely efficient job of delivering it in seconds. This magnitude and speed of delivery of information has organically pushed the consumption habits to rapidly change and evolve. Topics that used to trend for more than a week, just half a decade ago, can now change three times a day. This means that your consumer landscape is shifting exponentially per day, making old methods like consumer research, surveys, focus groups and other market research methodologies, obsolete. There is a need for brands to interact with consumers weekly, if not daily. None of the old methods can make that happen for us.

consumr.ai’s AI Twin technology enables you to create an AI persona feeding off the short- and long-term historical data points to give you a complete grasp of where the audience is headed.

2. Anyone can now talk to their consumers

Though we say that market research and focus groups are ancient methods, brands won't stop investing millions in them almost instantly. However, only a few professionals in a large enterprise will have access to organize a focus group or user interview with real consumers.

consumr.ai’s AI twin now democratizes these capabilities by enabling an almost immediate interaction and that too multiple times a day, to find answers to complex business problems within seconds.

3. Great replacement for focus groups

As we discussed earlier, focus groups are an integral part of receiving consumer feedback and using it to improve the product, sales, and after-sales services. However, it is not a hidden secret that the overheads to get the focus groups going are immense. Not to mention the sample size is poor and way below any minimum statistical confidence level. Due to a lack of alternatives, it was still the best thing to do. Top it up with the rate of speed of change now, which decreases its value significantly. Yet, it was the best possible way to get feedback. Social listening tools, to some extent, serve the purpose of receiving feedback directly from the customers, but they still have a lot of noise and lack perspective.

consumr.ai’s AI Twin adds perspective to extremely intelligent insights based on millions of data points at the speed of light. It's not time-bound, does not cost money for every run, available to as many users throughout an enterprise, and it's almost real-time. Mic drop…

How does it work?

An AI twin operates in a way that mirrors how an average human conscious mind functions. We have a collection of memories and experiences shaped by our personality. The model of an AI twin includes three essential components:

1. A persona: This is created by incorporating a behavior report. Keep in mind that these behavior insights are derived from first-party (anonymized) data, so the persona represents a cohort of your customers.

2. Memories: These are formed by adding two other insights from consumr.ai:

   1. Intent Report: This report informs the twin about what the audience is searching for and their mindset. It helps the twin form an intention regarding your brand and products.

   2. Mentions Report: This report provides nearly real-time information on what consumers are saying about your brand and products. It helps the twin understand the audience's market experiences.

3. Portfolio: Creating portfolios isn't a daily task, so you might overlook the importance of the drop-down menu at the top of consumr.ai that displays your organization name. This section defines aspects like product, brand, and category, which are passed to the AI twin references to ensure that all its conversations stay relevant to the product.

How to create an AI Twin?

The process of creating an AI twin is extremely simple. You need to start from the ‘manage’ screen. There are two very important prerequisites to create an AI twin:

1. You need to have access to all three modules, namely Behavior, Intent, and Mentions.

2. The behavior report should be created using an advanced AI feature, since it requires the persona to be created that is used by the AI twin. These personas are only created when the advanced AI feature is enabled.

To proceed, click on the gear icon at the top right corner after purple insight screen.

A panel will slide out from the right. In the top row below the heading called “management”, you will see a row that has two drop-downs. The first drop-down will be the organization. Most users will only have one organization assigned to them, but if you don't, then please select the organization where you want to create your AI twin. Once you've selected the organization, you can move to the second drop-down called “manage type”. This drop-down will be defaulted to “assets”. If not, please select it.

A section below the drop-down will appear with the organization name you have selected. In this section, there will be four tabs. Each tab will be for a type of asset. AI twin is a type of asset that will only grow in value the more time you invest in it. In the AI twin tab, if you already have a few personas (also known as AI twins) created, they will appear in the table. You can now choose if you want to edit an existing twin or create a new one. For now, let's create a new AI twin.

To the right top corner of the table, you will see a purple button that says add new. Click it.

A short form will instantly appear saying, “create new AI twin”. The first component out there is to select how many AI twins you want to create at once. In an ideal scenario, you would not need more than one AI twin. However, if you want to create more, you can select up to three AI twins that you can create at one go.

Next option is selecting a stage.

You will be creating a twin of your consumer to chat with. Hence, this drop-down becomes extremely critical. As you can see in the screenshot, there are six different stages available for you to select from. Each stage is a step in the consumer journey.

1. Potential:

This means if you want your AI twin to be a potential customer. This stage will help you understand how a consumer is thinking when he/she is in the market for your product category and may not have loyalty towards your brand.

2. Existing

This means if you want your AI twin to be an existing customer. Selecting this will help you ask questions to the customer who has the experience of using your product or brand.

3. About to purchase

Imagine a user who has crossed the top funnel of conversion and is now checking your prices, reading through the product page, either saving the product to buy later or adding it to a shopping cart. Knowing what this kind of consumer is thinking is incredible. Most feedback coming from this stage helps you understand and improve the customer experience of the user journey.

4. Newly acquired

This stage is a stage of extremes. This consumer paradigm would either be extremely positive or extremely negative. Either way, it would be amazing to understand how a newly acquired client would respond to your brand or product.

5. Long-time loyal

This is the customer that you would never want to lose. Hence, this increases the stakes. Staying abreast with what such customers want is critical to your churn or detraction. Understanding this persona's perspective is critical to the future of your sales.

6. Churned:

Last but not the least, churned are those sets of consumers that have stopped using your product and being your customer. They have either moved away from the category or moved to another brand. Knowing the perspective of this consumer can future-proof your brand and your product.

As mentioned above, you need to select a persona to generate the right behavior report. When a behavior report is created by enabling the Advanced AI feature, it generates an audience persona. AI twin uses that audience persona to build the personality of the twin. So, your persona is dependent on the behavior insight and the behavior insight is dependent on the audience you select. If you select a first-party audience, you can build an AI twin for your first-party audience.

The AI twin works on two critical aspects. One is the personality and the other one is the memory of the personality. So now, if you imagine a product or a brand of your business that you want to ask questions about, you will need to create a behavior report that will be used as the personality. Needless to say, that beyond the persona, the behavior report will be used for other behavioral data points as well. However, the intent and mentions report will be used as the memory of that persona. This means that selecting the right report becomes extremely critical. You cannot expect the AI twin to give you great answers if you select the wrong report. In fact, we strongly advise creating reports specifically to suit the use case of creating an AI twin.

Please note that, you cannot add more reports once the AI Twin is created. You can add multiple insights in the memory, but you need to select just one behavior insight per AI Twin. Also, once the memory is added to an AI Twin, there is no means to update the memory. To do so, you will need to create another AI Twin. Having said that, this is a feature we are working on and will build it in soon, when you will be able to auto update the memory of an AI Twin.

You cannot add file assets like .pdf, .pptx etc. to an AI Twin. At least, not yet.

Once you create an AI twin, it will be available to all the users of the portfolio. However, the conversation you have with the AI twin will stay private and only with you.

Once you create the AI twin, consumer.ai will generate a fictitious name for the persona and assign an AI-generated picture. The arrow you can see next to the profile picture allows you to change and influence the picture just to add a bit of fun factor to it with options like anime, cinematic, fantasy, sketch, pixel art, realistic and punk. Below that, you will get a summary of who the personality is. At the bottom of these details, there will be a button that says “converse with AI twin”. Click it to begin the conversation.

Conversing with an AI Twin

Name of the conversation: In the top left corner, you can see the name of the conversation. If you have not asked any question yet, the conversation is untitled.

Persona: To the right of the name of the conversation is the name of the personality and just above it is the mention that you are talking to an AI twin.

Stage: Further to the right, you can see the stage that you have selected for the personality, which in this case is “potential”.

In the far top right corner, there is a button with a drop-down menu that you can use to start a new conversation.

Left Panel: the left purple color panel is where all your options are with reference to this conversation and others.

• At the top of the purple pane, you will see the organization and the portfolio name. You will not be able to change them from here, but they are listed for your reference. To switch your portfolio or organization, you will have to exit the conversation modal by clicking the left-pointing arrow at the far left of the modal.

• Below that, you will see the AI twin properties. These properties are divided into two parts:

  • Persona: This is where all the behavior reports will appear. Per personality, you can only have one behavior report. You will be able to see, the behavior report that was used to shape the personality of the Ai Twin. You will also be able to open the insight in a new tab, should you want to see the report in detail.

  • Memory: Right below the persona, you will have mentions, intents, and/or other reports, and uploaded files like market reports or PDF files with consumer-related data in them. In other words, this is where you can see all the intent, mentions and market reports that are used in this conversation and also open the insight to view their details in a new tab.

  • Saved conversations: At the bottom of the left pane, you have a section where all your saved conversations will appear when you click it.

  

  Clicking on save conversation will open a second left pane that will list all the past conversations you have done within the portfolio. Click on it and the conversation will load in the rightmost pane.

The Conversation Pane

An AI twin is an expert in getting started. When you have all the assets ready, it will load a list of up to four questions to help you start the conversation. Keep in mind that these questions are just suggestions and may not be as relevant as your business problem. You can use them only when you can't think of what to ask. It is always recommended that you have a business problem in mind before you start conversing with the digital twin.

Start your conversation and you will be amazed by the answers it can provide. Below are some of the guidelines that you can use to converse with the twins.

What type of questions can I ask?

The questions that you should ask the AI twin are the second most important thing that you need to get right. This means that you should be prepared to ask the right questions. Below are some of the guidelines that you need to follow to ask the right questions to the twin:

1. The twins are based on highly complex data sets that come together to build a personality. Furthermore, a lot of other reports are used to build its memory. Before you ask business-related questions, it would be a good idea to ask it to tell you something about itself.

2. You can ask the twin for a perspective on what you think about a specific product of your brand.

3. You can ask the twin about how it uses your product based on its persona.

4. You can ask it to collate a list of use cases of your product.

5. You can ask what it likes and dislikes about your product.

6. You can ask it to draw a comparison between two products, provided it has the memory of both the products.

Other rules to help with better communication

1. Try not to ask two or three questions in one prompt. Ask individual questions, one at a time.

2. Do not use slang and ensure you explain your questions completely and in correct grammatical English. Though it has the capability of correcting a grammatical error and then interpreting the question, if your question does not make sense it may not be able to interpret what you're trying to ask.

3. Do not ask about a competitor or a product from your brand line for which a report has not been created and added to the memory.

4. Your questions should be clear not only in language but also in intent.

5. Remember, the digital twin is a complex set of algorithms that helps you get an answer to your question. By no means is it a vehicle to get your ultimate task done. For example: Do not throw questions like, "what should I target to get a ROAS of X%". It is intelligent but not as intelligent to replace you. At least not yet.

6. Do not upload the performance report of the campaign and expect it to answer your performance related questions, like: "What has been my weekly CPM for the last 10 weeks?"

7. Avoid getting too vague with your questions, like asking it about its personal life etc. Treat it like a representative of a cohort of your audience.

How is the different from ChatGPT or Claude?

1. Continuing from the previous example illustrating how AI Twin operates similarly to an average human conscious mind, a Large Language Model (LLM) like ChatGPT or Claude functions akin to your subconscious mind. These models are trained on vast amounts of internet data. When humans are asked a question, we don't immediately relay every piece of information we know; instead, our conscious mind retrieves pertinent details from the subconscious and refines them based on the context to provide a succinct response.

2. The distinction between spontaneously sharing information and delivering carefully considered details is significant. Imagine you're preparing for a training session or an important presentation before senior management; it is improbable that you would improvise entirely in such circumstances. You would likely prepare days in advance, organizing a structured flow of information, compiling a knowledge base from what you already know, and researching areas where you lack information.

3. Thus, when you ask the same questions to an AI Twin as opposed to an LLM like ChatGPT, you can expect more relevant, well-structured, and thoroughly researched responses from the AI Twin.

4. An AI Twin's scope is confined to interacting with the consumers of your product or service and serving as a consumer itself. Approach your interactions with the AI Twin as a “conversation” rather than an “interrogation.” AIs are typically designed to be agreeable and polite, so they won't respond with doubt, disrespect, or aggression. They generally aim to accommodate rather than challenge you. This aspect is something that the leading experts at ProfitWheel and other AI companies worldwide are working to improve.

5. While AI Twin excels in logical thinking and reasoning, it should not be expected to replace a human entirely.

It's Limitations

AI represents the latest major technological innovation, and as such, many aspects still require refinement both technically and ethically. Consequently, for legal, controversial, and cautionary reasons, AI tends to be more restrained in its responses and will not engage in conversations as a human might. When challenged, it often yields to the user's perspective. Therefore, effective interaction with AI necessitates certain safeguards. If you encounter an issue where the AI fails, it is advisable to rephrase the question and ask again. Instead of confronting it, encourage the AI to reassess the data or highlight inconsistencies in its previous comments.

Additionally, it is beneficial to avoid using slang or words with multiple meanings. For example, during testing, I queried the AI about the Samsung Galaxy Note by asking, "What do you like about your note?" Here, the term "note" holds various interpretations, and the context was unclear, resulting in an irrelevant yet portfolio-accurate response.