Step 1: Define Your Sign-Off Policy Before Configuring Agents
Most organisations don't lose grants because they wrote a weak application. They lose them because someone hit submit before the budget was approved, or because a key partner wasn't looped into the final version. Before you let any AI agent touch your pipeline, you need a clear answer to one question: who in your team is allowed to sign off a submission?
Autopilot vs assisted: two modes, two levels of control
GrantGunner’s AI agents operate on a spectrum. In assisted mode, the agent drafts, gathers evidence, and flags missing information - but it stops and asks for your approval before anything goes out the door. This works well when you have a named grant lead or committee that needs to see every application.
In autopilot mode, the agent progresses applications through to submission on your behalf, only pausing when critical information is missing (for example, a mandatory financial statement you haven’t uploaded). The product won’t guess or fabricate data - but once it has what it needs, it will submit unless you tell it otherwise.
Your sign-off policy determines which mode is appropriate for which opportunity. A small £500 community grant might be safe on autopilot. A six-figure Innovate UK bid? You’ll want full human review.
Write the rule, then configure the agent
Map your decision to GrantGunner’s settings: for high-value or sensitive applications, require the agent to wait for manual sign-off before submission. For lower-stakes, time-sensitive bids, set it to proceed if all required fields are complete. A clear policy means the agent knows exactly when to pause - and when to push forward.
Step 2: Set Agent Permissions in the Workspace
Where to find agent permissions
Open the Workspace for any live application and look for the Agent Controls panel 14usually near the top of the page, next to the opportunity name. You 19ll see a toggle for each agent type: 1cAI Scout for discovery, and drafting agent for form-filling. Click through to the permissions page, and you re in the right place.
Three permission levels to choose from
GrantGunner lets you dial the agent 19s role up or down in the same workspace. The key options:
- Draft only the agent fills in form fields, attaches evidence from your uploaded documents, and stops. No pipeline stages are moved, and nothing is submitted without you. This is the safest starting point if youre new to AI agents.
- Auto-progress to sign-off gate the agent moves the application through stages (for example from processing to in review ) but holds the final submission until you approve. Youll get a notification and a Review and Release button in the workspace.
- Full autopilot (still with cancel window) the agent drafts, progresses, and submits. Even here, the product builds in a short window before the submission is finalised so you can pull it back. Useful for low-stakes or recurring applications where the content barely changes.
Trust underpins your permissions
When you grant any level of access, your profile data and uploaded documents are used only to complete that application. They are not shared with other users or used to train third-party AI modelsa point GrantGunners FAQ-style copy is explicit about. That matters because you are effectively asking the system to handle sensitive budgets, partner letters, and financial statements. Knowing those materials stay inside your account makes it easier to give the agent enough room to actually save you time.
Step 3: Customise the Pipeline With Review-Stage Gates
The default GrantGunner pipeline - Processing → In Review → Submitting → Submitted - works fine for solo founders. But if you’re part of a charity board, a university research office, or a multi-stakeholder social enterprise, you need more control.
Add custom stages for your approval process
Open the Pipeline Settings from your dashboard. You’ll see an option to Add Stage. Name it something that matches your real workflow: Awaiting internal review, Director sign-off required, or Budget check. Drop it between the drafting stage and the submission gate.
Each custom stage can block automatic submission. Toggle Require human confirmation before moving on for that stage. Now the AI can draft, gather evidence, and populate forms - but it cannot push the application past your review stage until a person clicks ‘Approve’.
Connect custom stages to the calendar
One risk: applications stall in review and you miss the deadline. GrantGunner’s Calendar solves this. Every stage you add inherits the opportunity’s deadline. Even if an application is stuck at Awaiting Director sign-off, the deadline still shows in red on your calendar. You get a warning days before the close - not after.
Set a reminder trigger per stage too. Right-click any custom stage, choose Set notification, and enter a lead time like ‘3 days before deadline’. The platform alerts you: ‘Application at Director sign-off - deadline in 72 hours.’ No more discovering a held submission the day after the window closed.
Example in practice: A community sports club with a £50,000 grant application sets up three extra stages: Trustee review, Budget sign-off, and Final Director check. Each requires a named person to confirm. The AI drafts the full bid, but it sits at Budget sign-off until the treasurer approves the costings. Meanwhile, the calendar shows the deadline approaching. The club chair gets a notification and chases the treasurer - all before the due date. The application submits on time, with every signature in place.
Step 4: Test With a Live Opportunity and Audit the Output
Pick a low-stakes opportunity and run it end to end
Before you trust your grant pipeline to live agents, run a controlled test. Choose an application you care about least - maybe a small trust grant or a local council fund under £5,000. Nothing that would destabilise your charity’s core budget if something goes wrong.
Assign one team member as approver and another as writer. The writer configures Agent Controls as you defined them in Step 2: autopilot drafting on, but submission paused at the sign-off stage. The approver watches from the pipeline and waits for the “Awaiting approval” notification.
What to check in the test run
Watch the agent’s output closely on these three points:
- Draft quality: does the AI pull the right context from your profile and past applications? If it misstates your mission or budget figures, tighten your source documents in the workspace before going live.
- Information requests: GrantGunner’s agents will pause and ask you when critical information is missing. Count those interruptions. Too many? That signals gaps in your profile that you should fill now, not mid-deadline.
- Sign-off gate triggers: when the writer finishes, does the application land in the correct pipeline stage (e.g. “In Review” or your custom “Awaiting Director Sign-Off”)? And once the approver clicks approve, does it proceed to “Submitting” automatically? If not, revisit Step 3’s stage rules.
Run one full cycle - draft, review, approve, submit. Fix any friction you spot. Then you’re ready to turn the agents loose on real deadlines.