The use of agentic AI for A/B testing is literally a game-changer for marketing teams, making it much easier to scale testing and experimentation programs. But with some companies making bold claims about what their AI can do, it can be hard to know just what to believe. So let’s look at how the competition’s AI really stacks up against ours.
Move at the speed of evidence with Evi
First, let’s look at our agentic AI. Launched in November 2025, Evi is AB Tasty’s AI-powered marketing agent designed for evidence-based decision making. It transforms complex data into clear, actionable strategies for repeatable, measurable results and ensures every step you take is grounded in evidence.
But Evi isn’t just one tool. Evi is a suite of intelligent AI agents integrated throughout the entire AB Tasty platform, all optimized for specific tasks.
| Agent | Function | Description |
| Evi Ideas | Ideation | Scans pages and generates data-backed ideas for new tests based on visual and contextual input. It uses AB Tasty’s proprietary data and UX principles. |
| Evi Hypothesize | Hypothesis creation | Uses a checklist of essential elements to help you create well-structured hypotheses with clear objectives. Assigns quality scores, highlights gaps and suggests edits. |
| Evi Content | Visual editor | Turns natural language prompts into precise on-page edits (HTML/CSS/JS) with no coding required. |
| Evi Analysis | Post-test analysis | Analyzes campaign data, delivers clear, actionable insights. Highlights winning variations and breaks down why they drive transactions. |
| Evi Feedback | Qualitative analysis | Analyses Net Promoter Score (NPS) and Customer Satisfaction (CSAT) feedback. Quickly identifies key themes and provides actionable insights from customer comments. |
| Evi Explore | Revenue insights | Powered by our patented metric, RevenueIQ, provides real revenue projections per visitor / per month with confidence intervals. Let’s you see what each test is worth before you launch. |
| Evi Formula | Catalog attributes (R&M) | Self-serve tool for creating catalog attributes using natural language prompts. |
So how does Evi compare to the AI used by our main competitors?
Evi vs. Optimizely Opal
Opal is the name of Optimizely’s suite of AI tools integrated throughout their platform. It’s not a standalone product, but rather a collection of different AI agents threaded across their entire product suite, which, along with Experimentation, also includes CMS, CDP, and Commerce.
Indeed, most of Opal’s AI agents are actually focused on CMS, CDP, and Commerce rather than Experimentation. One potential drawback for customers is that many of these AI features are tied to using the entire Optimizely tech stack. Rather than talk about Opal’s features for other areas, let’s look at what AI features Opal does have specifically for Experimentation:
| Feature | Description |
| Test ideation | Generates ideas for new experiments based on URLs and brand tone. |
| Variation editor | AI-assisted creation of test variations based on Google Gemini. |
| Campaign creation | Creates containers for both web and feature experimentation. |
| Variable suggestions | Suggests flag variables and variations in feature experimentation. |
| Chat-based data exploration | Allows conversational exploration of test data. |
| Results summarization | Summarizes test results and provides directional guidance. |
| Experiment advisor agents | These include a personalization advisor, experiment planner, and results summarizer. |
| Experiment scorecard | Scores experiments from the analytics interface. |
Opal’s AI agents that are used specifically for testing and experimentation are very comparable to those of Evi. Both have dedicated AI agents for ideation, editing, and analysis. However, Evi also includes our proprietary RevenueIQ analysis and can leverage AB Tasty’s other AI features, EmotionsAI and Wandz for targeting and segmentation.
Some of Opal’s features also appear to be standard statistics features that have been rebranded as AI (e.g. multi-armed bandits and sequential testing).
Key differences between AB Tasty’s Evi and Optimizely Opal
- Price: Evi’s AI features are included in all contracts at no additional cost. Opal is a paid add-on that costs around US$30,000 extra.
- Speed: Evi’s AI editor is based on OpenAI’s ChatGPT and proven to be faster than that of Opal, based on Google Gemini.
- AI Targeting: Evi can leverage our other AI features, EmotionsAI and predictive targeting (Wandz) for AI-based segmentation. Opal has nothing comparable.
- Revenue Analysis: Evi Explore is based on our patented RevenueIQ metric for ROI projections. Again, Opal has no equivalent.
- Experimentation focus: Evi is 100% focused on testing and experimentation. Most of Opals AI agents are designed for CMS/CDP/Commerce.
Evi vs. Kameleoon PBX
Kameleoon PBX (Prompt-Based Experimentation) is an AI-powered tool that allows users to generate A/B tests directly from natural language prompts. It is positioned as an all-in-one AI agent for test generation, fully integrated with Contentsquare.
Here is a list of PBX’s key AI features:
| Feature | Description |
| Prompt-based test generation | Users write prompts describing what they want to test, PBX then generates the necessary code/variation. |
| Contentsquare integration | Turns behavioral insights from Contentsquare into A/B tests. |
| Automatic site analysis | Upon URL integration, 3 to 4 AI agents analyze the site’s structure (HTML/CSS/JS) giving prompts strong contextual awareness. |
| Figma integration | (Currently a Beta feature) Allows the uploading of mockups from Figma, reducing errors in banner creation and saving time in implementation. |
| Code generation | Generates HTML, CSS, and JS code for experiments. |
Unlike Evi’s suite of specialized AI agents, PBX is a single generalist AI agent. This provides you with a limited amount of control and can make it hard to iterate. Kameleoon claims that by using PBX specifically, its customers can build tests faster, more accurately, and at less cost per test. But the reality is that these improvements aren’t specific to PBX, all vendors with agentic AI see similar positive impacts for their customers.
Key differences between AB Tasty’s Evi and Kameleoon PBX
- Price: Evi’s AI features are included in all contracts at no additional cost. Like Opal, PBX is a paid add-on.
- Usage: All AB Tasty customers can make unlimited use of Evi, while use of PBX is based around credit quotas.
- Architecture: Evi is a suite of different specialized AI agents. PBX is a single generalist agent.
- Speed: Evi has a prompt response time of around 30 seconds, compared to up to 3 to 4 minutes for PBX.
- Advanced segmentation: Evi can leverage AB Tasty’s other AI features like EmotionsAI for advanced segmentation. However, like Opal, PBX has nothing comparable.
- Structured output: Evi supports structured output (JSON, rollback, versioning), while PBX makes no mention of whether this is the case.
- Production quality: Evi is fully production-ready, while some customers have reported that PBX has QA issues and webperf impact.
Full comparison: Evi vs. Opal vs. PBX
Architecture, Pricing, Performance
| Criteria | AB Tasty Evi | Optimizely Opal | Kameleoon PBX |
| AI architecture | Multi-agent system | Suite of AI features across platform | Single generalist AI agent |
| Agent specialization | Task-optimized agents (Ideas, Content, Analysis) | General-purpose AI tools | One-size-fits-all, prompt based |
| Structured output | JSON structure, rollback, versioning | Limited | No mention |
| Philosophy | Evidence-based, grounded in proprietary data | Platform-wide AI integration | Prompt-to-test generation |
| Included in price | Yes, all contracts | Paid add-on (~US$30K) | Paid add-on (+25% list price) |
| Usage model | Unlimited | Credit-based | Credit-based/quotas |
| Availability | All users | Requires purchase | Requires purchase |
| Prompt response time | ~30 seconds | Reported as slow | ~3 to 4 minutes |
| Production readiness | Yes | Yes | Demo-ready |
| Code quality | Optimized HTML/CSS/JS | Variable | QA issues reported |
Feature comparison
| Feature | AB Tasty Evi | Optimizely Opal | Kameleoon PBX |
| Test ideation | Evi Ideas | Test ideation | Limited |
| Visual editor | Evi Content | Variation editor (Google Gemini) | Prompt-based generation |
| Hypothesis creation | Evi Hypothesize | Natural language interface | Not mentioned |
| Results analysis | Evi Analysis | Results summarization | Not mentioned |
| Revenue projections | Evi Explore (RevenueIQ) | No equivalent | No equivalent |
| Feedback analysis | Evi Feedback (NPS/CSAT) | Not mentioned | Not mentioned |
| AI-based targeting | EmotionsAI/Wandz | No equivalent | No equivalent |
| Contentsquare integration | Not native | Not native | Native |
| Figma integration | Not mentioned | Not mentioned | Beta feature |
| SPA/Dynamic JS support | Yes | Yes | Yes |
