June 5th 2025
The following stories will go live on the app this June 5th, 2025.
1
Creative Evaluation – New Creative Variation
2
Product Category Stage of AI Twins
3
Guided Google Ad Campaigns
4
Auto Mode Conversations
5
Investigative Meetings
June 5th’s release will help you flex your muscles in front of your clients. Now that we have built the foundation, we are adding a higher degree of intelligence to the product. Many of you may have already heard of what is coming up. However, here is a formal communication of what you will be able to experience on June 5th, 2025 by 9 am ET.
Creative Evaluation
In our meeting module, we have had Creative Evaluation for some time now. However, given the feedback that we've received from clients, we have realised the potential and use cases — and hence, this release brings a new beta feature that adds actionability to Creative Evaluation.
The problem that we discovered with the current version of Creative Evaluation was that even though the output was insightful, there was little that a creative resource could do to convert text into creative output.
So, we used the textual feedback and fit it into a model to create modifications to the current creative that is being evaluated. Yes, the new variations may not be totally flawless, but they would still be able to appeal to a creative resource on how the feedback can be incorporated.
To use it, start a new meeting and select Creative Evaluation. Select Ad Image and upload a creative you want to evaluate, then hit the submit button.
It will give you an option to either select available AI Twins manually or you can hit the Recommend Twins button to auto-select.
After you select the Twins, the system will create four to five questions to ask these Twins about the creative. These questions will be custom-made.
You can begin the meeting, and it might take up to 5 minutes to produce the results.
The result will give you pointers on what the AI Twins liked, what they did not like, and what can be done to improve the creative.
Right at the top of the page, you will see an option to create a variation that will take the feedback of the Twins into account and modify the image to create a variation incorporating all the feedback.
You can create multiple variations — up to 5 — and also download them to share with others or upload into a campaign.
In the next release scheduled for June 17th, 2025, we will be launching the second phase of Creative Evaluation that will enable the system to pull creatives from a live campaign, assess them based on performance and AI Twin feedback, and improve them accordingly. We will keep you updated about it through another release note.
Product Category Stage
By now, we're all aware that conversion stages for every Twin add a unique perspective based on the stage it is in. However, we have time and again come across clients who stated they wanted the Twin to not be biased by the brand or the product. What if the client wants to talk to a consumer who is not even aware of their brand? And what if a user wants to conduct market research for a product that the client has not built, but variants of such products exist in the market?
Since product category can create maximum impact in the online environment, we have added this feature to it.
To incorporate this stage, we have brought in 3 changes:
Add Category input to Omnibox
You can see categories show up in the tab above the Omnibox. Users can type in a minimum of 3 letters to get recommendations, and then execute a report using Categories.
Make changes to the Mentions report when category input is used
Add Product Category as a stage of conversion while creating AI Twins
The process of creating a Category Twin is like creating any other AI Twin. If you are selecting existing assets, make sure you create Category assets from the Omnibox or that there are existing assets made from Category as inputs. If you are looking to create new assets, then the recommendations that you get will be broad and category- or industry-level — without any hint of a brand.
Guided Google Ads Campaigns
The addition of brief in the Omnibox and AI Twin Creator got us some great feedback, since the recommendations it was able to come back with were brilliant. Hence, we decided to add 2 features related to this in building Google Adwords — PMax and DV360 campaigns.
View how it works, here: https://app.birdeatsbug.com/PZyuXUOvG6eetb-pYIz0EBl9tN-UdaqGhJu2ClI1DGK2
Steps to follow:
Create a new campaign and select Google and Performance Max (or DV360).
Select a campaign type except Affinity Segment.
In Campaign Presets, it will ask you to select a flow type. If you select Manual, you follow the regular process. But if you select Guided, then a multiline textbox appears asking you to enter a brief. You can enter a brief the same way you would in the Omnibox or AI Twin Creator.
Once you type in your brief, the system will select an objective for you. You will have the option of changing the objective.
Give your campaign a name or accept the recommended name.
In Signals, you can choose to enter Keywords, Videos, and URLs — any two of them or all three — and hit the Get Targeting Recommendations button.
It will give you a max of 6 recommended keywords. You can accept them all, replace them, or search for keywords you want. Remember, you cannot select more than 6 keywords.
You can then create a plan. It will take up to a minute.
Once the plan is created, you will see a regular word cloud and a table where all the keywords will be listed. All the keywords will be preselected by default. The user can remove as needed.
The second change we have brought about is to refine the plan. Here, we have given users the option to add the influence of a Behavior Asset or an AI Twin. Selecting an AI Twin will look at the original plan of hundreds of keywords and then shortlist about 50% or fewer based on the influence of the AI Twin. The same applies when you select a Behavior Asset — it uses the persona of the Behavior Asset and applies that influence over the keywords to refine them.
You can either download or activate the campaign.
In summary: We can now build campaigns using a brief to come up with keywords and execute campaign targeting. Once the targeting is ready, you can refine it by adding the influence of an AI Twin or a Behavior Asset.
Auto Mode Conversations
Over the last 18 months, we have made multiple modifications to conversations. Every time, the problem statement was that users were not aware of what to ask a Twin. We’ve been monitoring the kind of questions users ask — and most of them are either unclear or treated like a Google search.
We’ve been working on various ways of recommending questions as thought starters.
In a recent experiment, we were able to look at an objective a user would like to achieve — and come up with extremely intelligent questions. To put that intelligence to good use, we have come up with Auto Mode Conversations.
While conversing with an AI Twin, please note the robot icon in the text box at the bottom of the page. When you click it, it enables Auto Mode.
A dropdown of objectives will appear for the user to choose from.
The next control is a slider that allows the user to select how many questions he/she would like the system to generate and answer.
We have put a restriction of maximum 5 questions, after which the user can continue the conversation by asking follow-up questions.
If the AI Twins end up meeting the objective before the selected number of questions, the system will not ask the remaining ones.
What’s extremely interesting is that each question is progressively created based on the AI Twin’s last answer. This ensures that the questions are not repeated — and the answers are unique.
Adding an objective to the conversation ensures that it is heading in a direction to meet the goal we’ve aimed for. This is what makes the entire conversation extremely intelligent, current, and helps prove a point to the client.
Investigative Meetings
Since we were able to crack the problem of asking intelligent and current questions, we thought of repurposing it in a new type of meeting — called Investigative Meetings — that executes a meeting based on auto-generated intelligent questions discussed by a group of Twins.
🎥 Here’s a recording of how you can initiate an investigative meet: https://app.birdeatsbug.com/u52GVTzQLyHcJ7UxmZD6z6u3Jtr2GbwstZNsapblnF-q
Once the meeting is completed, a report and summary of the meeting is provided. Our next intent is to enable follow-ups for every meeting — allowing users to converse with the intelligence and findings of the session.
Last updated