Bïrch, the platform associated with the former Revealbot name, should be evaluated using its current product surfaces. Its public documentation now includes AI context and MCP workflows alongside condition-based rules. An old comparison that calls it only a manual rules engine misses that change.
This guide is based on public first-party material checked September 7, 2026. GaaS publishes advertising software, and we have not performed a hands-on Bïrch account test for this article. The exercises are proposed evaluation tasks, not findings about actual reliability.
Separate the rules engine from AI-assisted authoring
The Bïrch features page describes nested conditions, metric comparisons, ranking conditions, custom metrics, and automation logs among its capabilities. These are mechanisms for specifying and inspecting recurring account work.
The Bïrch AI page describes business context, access to account and rule information, performance analysis, and creating draft rules that users can review before activation. It also marks some capabilities as coming soon; do not count roadmap items as delivered functions.
Assess both layers. A well-written rule can run reliably while expressing a poor business decision. A useful AI suggestion still needs to compile into the condition and action the buyer actually intended.
Bring a rule with an intentional exception
Use a realistic policy such as flagging campaigns with deteriorating mature performance, while excluding an active experiment or constrained launch. Define the metric, reporting window, minimum evidence, target objects, and intended action.
Ask the AI-assisted workflow to draft the rule, then inspect the resulting conditions. Check whether the exception remains explicit rather than disappearing into a broad name filter or a simplified threshold.
Use read-only analysis or an inactive draft for this stage. The evaluation should reveal ambiguity before any live rule is activated, within the access and approval boundaries your organization authorizes.
Inspect how conditions compose
Test AND and OR grouping, object scope, time windows, and the treatment of missing values. A small logical difference can select a very different set of campaigns.
An illustrative case is a campaign with low reported ROAS but an incomplete conversion window. Ask how the rule distinguishes genuinely mature underperformance from delayed reporting. Do not assume a popular template accounts for your sales cycle.
Check whether custom metrics use consistent currencies, revenue definitions, and denominators. A rule that compares two similarly named but differently defined metrics can produce technically valid actions with weak commercial meaning.
Test the AI context as a maintained record
Provide a short current business constraint, then revise it in a controlled draft exercise. Ask the workflow to explain which version it is using and whether an existing draft needs review after the change.
The aim is to understand how the team will maintain context, not to prove that one prompt makes the system permanently aware of every business change. Product launches, seasonal offers, and new client limits need owners and update processes.
Use the conflicting-instruction guide to prepare a case where a broad performance goal conflicts with a specific account restriction. Record how the demonstrated workflow resolves or surfaces that conflict.
Examine activation, repeated execution, and logs
Before activating an authorized rule, inspect the selected objects and exact actions. Ask what happens on subsequent checks, after a partial failure, and after a manual change by another operator.
A useful evaluation worksheet includes:
| Scenario | Question to resolve |
|---|---|
| Same condition remains true | Can the action repeat, and within what limits? |
| Data source is delayed | Is the missing or stale state visible? |
| Another rule touches the object | How are conflicts inspected and controlled? |
| One action fails | Is the unresolved item identifiable? |
| Rule is disabled | How is cessation verified? |
These are questions for the actual product demonstration. Public feature listings alone do not establish the answers for every rule type or channel.
Evaluate MCP scope explicitly
Bïrch's current navigation advertises MCP connections and AI workflows. Ask which connected assistant can read data, create drafts, or execute actions, and how account scope and user permissions apply.
Do not infer that a connected chat client has the same controls as the main product interface. Inspect the exact integration your team intends to use, including approval behavior and how actions are recorded.
Keep client data and rule libraries scoped appropriately in an agency setup. A reusable process can be shared without copying another client's private economics or account instructions into a new workspace.
Include maintenance in the comparison
Record the time spent designing, reviewing, debugging, and updating rules. A workflow that saves daily clicks may still require meaningful maintenance when account structure, naming, conversion definitions, or platform behavior changes.
Ask for current plan and channel coverage, user permissions, billing basis, and access to logs or exports. Avoid assuming every advertised surface is included in every plan or has feature parity across providers.
The AI buyer versus rules-engine guide can help identify which tasks benefit from explicit conditions and which require broader investigation. Many teams need a combination rather than a single category label.
Conclude with task-level evidence
Bïrch may fit a team that needs repeatable condition logic and wants AI help analyzing or drafting that work. That is an inference from the documented workflow, not a claim of superior results.
Use the pilot scorecard to separate demonstrated behavior, documented claims, unresolved questions, and business outcomes. The strongest conclusion explains whether the tested rule and AI workflow improved your operating process under the actual account conditions.