My colleagues and I at Newgens are developing Action! Lab, a workspace for handling everyday business requests. We want to help teams turn information from messages, spreadsheets and screenshots into clear tasks, then review the details and keep track of what happens next.
What is Action! Lab for?
A request rarely arrives with every detail in the right place. Someone sends a message, another person adds a spreadsheet, and a missing detail turns up later. Before work can begin, somebody has to bring that information together.
Action! Lab helps organise the input, prepare a structured draft and identify what still needs checking. The person handling the task can review the source, fill in missing information and keep the result with its task record.
Is it an agent, a workflow or a data tool?
I would describe it as a business workspace that combines agents and workflows. An agent helps interpret a request and work through unclear information. A workflow connects the steps that follow agreed rules. SQL-style queries are supporting data tools underneath that process.
The agent part runs an Agent Loop: understand the current information, choose a useful next action, inspect the result and adjust. The workflow keeps agreed steps and review points in place. This lets us use judgement where the task is unclear and predictable processing where the rules are already known.
For the user, the useful thing is the task itself: what information has arrived, what needs attention and what can happen next. They should be able to work through that without having to choose which technology handles each step.
What can it help a team do?
Take a staffing request. A colleague receives a message asking for ten teachers, with a location and working hours. Action! Lab can help turn that into a job-requirement draft, retain the requested headcount and flag an unspecified salary or start date for confirmation.
The same approach is useful for organising incoming service requests, checking whether a form is complete and bringing task information together for review. The aim is to reduce repeated copying, make missing details easier to spot and leave a record that someone else can follow.
We are starting with information intake, draft review and task records, then extending the workflows around them. What matters to us is whether the person doing the work can understand the result, correct it and decide what should happen next.
The concepts guiding our next stage
In an InfoQ interview, the idea of a thin Agent Loop and a substantial Control Plane helped me put several product decisions in context. The loop handles the agent’s immediate work. The Control Plane is the part responsible for task state, permissions, confirmations and recovery. For Action! Lab, that means the work should remain understandable and manageable as agent capabilities develop.
Workspace and Checkpoint are two practical concepts here. A Workspace brings a task’s materials and intermediate results together, beyond a single Session. A Checkpoint records a verified stage of progress. We want these to help users resume an interrupted task from a result they can trust, rather than reconstruct everything themselves.
The interview also argues for simple collaboration through explicit Input/Output. We will add parallel roles where separate work is useful, with a clear responsibility and result for each agent. A Declarative Goal means the user states the required outcome and constraints; the system can choose suitable steps without making the user manage every agent interaction.
Fail Fast and Failover describe the recovery behaviour we want to build. Fail Fast stops an invalid step before its error spreads. Failover lets another execution take over from reliable saved state. A brief network problem may justify a limited retry; an action with an unknown outcome should first be checked so it is not performed twice.
These are directions for the next stage of Action! Lab. We will judge them through everyday cases: can a colleague pick up a paused task, inspect what has been completed and continue safely?
Related product
The Action! Lab product section shows the workspace and its task views. If your team has a process that involves repeated information gathering and checking, you can bring that example to a product demo.
View Action! Lab