AI Agents and Assistants
What is an AI agent? A plain-English guide for business owners
What an AI agent is, how it works, where it helps a small business, the risks, and a 7-point check to see if you are ready for one.
Oversight
Human Approval Workflow
Every AI-proposed action stops at a standing approval gate before it takes effect, so a person always signs off first.
You want AI to speed up the work, but you are not ready to let it act on its own. This workflow puts a standing checkpoint in front of every AI-proposed action. A named reviewer sees the proposal in Slack, Teams or email, then approves, edits or rejects it before anything takes effect. Every decision is logged in your own systems, with who made it and when.
- Nothing the AI proposes takes effect without a person's sign-off
- Approve, edit or reject in Slack, Teams or email
- Full decision history for audits and accountability
$2,400
20–30 working days
Optional maintenance $249/mo
What it is
Human Approval Workflow is a standalone approval gate for AI tools we did not build. Examples are a third-party agent or an in-house script. The AI proposes an action, and a named reviewer approves, edits or rejects it before anything takes effect. For an Upmark Labs product, the Approval Workflow Layer add-on does this instead.
Why businesses buy it
Many teams stall on AI because one wrong automated action, such as a refund sent in error or a record overwritten, feels too costly to risk. So the work stays manual and slow. This workflow lets the AI do the drafting and preparation while a person keeps the final say on every action. You get the speed without giving up control, and a clear record of every decision for when someone asks why. The scope is fixed in writing before you order: one workflow, the connected systems and the approval rules, so the build has clear edges.
Problems this solves
AI actions take effect before anyone has checked they are right
Every action the AI proposes stops at a fixed checkpoint. A named reviewer approves, edits or rejects it, and nothing reaches your connected systems until that decision is made.
Approvals happen over email and chat, so requests get lost
The workflow gives each proposal one route to its reviewer, built across up to three connected systems. Nobody chases messages, because the proposal, the decision and the outcome follow the same path.
Leaders hold back from using AI on sensitive tasks
Because the gate is permanent rather than occasional, AI can draft or decide inside a single workflow of up to twelve steps. A person still signs off before anything takes effect.
Reviewers redo AI suggestions from scratch because they cannot edit them
Reviewers can edit a proposal directly instead of rejecting it and starting again. Twenty test runs on your own sample data check the approve, edit and reject paths before go-live.
A good fit if
You want every AI-proposed action reviewed by a person before it goes live, with no exceptions built into the gate.
Your process currently lets an AI act on its own, and you want a standing checkpoint added without redesigning the whole workflow.
You need a clear, permanent record of who approved, edited or rejected each action, for accountability.
Your reviewers are comfortable checking every proposed action, not just the ones an AI flags as uncertain.
Not the right fit if
The AI you want to supervise is an Upmark Labs product, bought now or earlier. Add the Approval Workflow Layer add-on to that product instead.
You only want a person pulled in for specific situations, such as low confidence or high risk, not every single action. See AI Escalation System, our rules-based conditional build.
You want an AI that actively supervises other agents' ongoing work and can intervene on its own. See AI Supervisor Agent.
You need multiple agents coordinated in one workflow rather than a single approval gate. See AI Agent Orchestration System.
Your workflow needs more than 12 steps or more than one AI decision point. Use our Custom AI enquiry form instead.
How it works
- We map your current process and identify exactly where the AI proposes an action, so the approval gate sits at the right point, not just somewhere convenient.
- A review queue is built showing each proposed action with its context, so your reviewer can approve, edit or reject it without hunting for background information.
- Rules define what the AI may draft on its own and what always needs a person, agreed with you rather than left to a generic default.
- The gate connects to the tools your reviewers already use, such as Slack, Microsoft Teams, email, your helpdesk or your CRM.
- Every decision, approve, edit or reject, is logged with who made it and when, building a full history for accountability.
- We run 20 test cases on your sample data before handoff, then deliver a recorded session and an admin guide covering how to adjust the gate.
What you provide within 72 hours
Your portal lists these the moment you order. Delivery starts the working day after everything is in and your access is checked.
- Your business information (company basics, hours and brand voice), shared during intake, or confirmation we can reuse what you gave us on an earlier order.
- The systems involved in the workflow, with admin or API access for each.
- The AI tool or agent the gate will sit in front of, and how its actions reach your systems today.
- Five to ten real actions the AI has proposed, so the review queue reflects genuine cases.
- Who should review each proposed action, and who receives alerts or errors.
- The rules for what the AI may draft alone and what always needs approval.
- Admin access and API keys for every connected system, shared by adding our team as a user or creating a key you control, never by email.
- One decision-maker who can approve the final gate rules and routing.
- An API key for your own AI model account (for example OpenAI or Google Gemini), created for this build with a spending limit.
The 72-hour rule. Send everything on your checklist, including access, within 72 hours of ordering so your build starts on time.
Timeline
Start
business information confirmed and the approval point in your process agreed.
Weeks 1–2
Review queue and approval gate built and connected to your named systems.
Week 3
Reviewer routing and the approve, edit or reject rules configured.
Weeks 3–4
20 test runs completed on your sample data.
Week 4
Your review and two rounds of changes to the gate or routing.
Go-live
Live, recorded handoff session, and the 30 days of support begins.
| Option | Delivery time | Price |
|---|---|---|
| Standard | 20–30 working days | $2,400 |
| Express | 15–25 working days | $2,900 |
| Rush | 10–20 working days | $3,350 |
What you receive at handover
- A single approval workflow of up to twelve steps, running in the background across up to three of your systems
- Reviewer routing, prompts and guardrails set up for the AI drafting or decision step
- A test report from twenty runs on your own sample data
- An admin guide plus the recorded handoff session
- 30 days of post-launch support
Expected benefits and how to measure them
- Your team can use AI drafting while a person stays responsible for each action.
- Helps cut back-and-forth over email, since every proposal follows the same route.
- Gives reviewers one consistent way to approve, edit or reject each item.
- Approval turnaround
- Compare request-to-decision times in your current process with the timestamps in the new workflow after launch.
- Edit and reject rate
- Count how many AI proposals reviewers edit or reject each week, to see whether drafts improve over time.
- Errors reaching your systems
- Compare after-the-fact corrections in the month before launch with the month after.
Before you start: Time how long approvals take today and note how they are requested. Write down recent mistakes that slipped through without a check.
These are things to track, not promised results. Outcomes depend on your data, your team and your market.
Running costs
The build price is one-time. Running the workflow afterwards costs roughly $20 to $100 a month in your own accounts, covering the automation platform plus AI model usage for drafting each proposed action.
AI model usage during build and testing is also billed to your own accounts, so ownership of every provider relationship stays with your team.
Ask about running costsCustom requirements
If your workflow needs more than 12 steps, more than one AI decision point, or more than three connected systems, add a note at checkout. You get a written Change Quote, and your original order keeps its normal timeline. Larger or bespoke projects use the Custom AI request form instead.
Request a custom versionKeep it running
We recommend monthly maintenance, whose change hours cover adjusting a reviewer, a routing rule or the workflow steps as your process evolves. Maintenance starts after the 30 days of support and can be cancelled any month.
$249/mo (the minimum fee, includes 2 change hours)
Optional. Starts after the post-launch support period. Cancel any month.
How maintenance worksHow this compares
| Do it yourself with a SaaS tool | Hire a custom agency | Upmark Labs | |
|---|---|---|---|
| Price | An approval-routing tool charges a per-seat monthly fee | A custom workflow build is quoted after discovery | Ours is one fixed price, shown on this page before you buy |
| Scope | A routing tool leaves the gate logic for you to configure | An agency agrees scope during the project | Ours is written to your exact process before you order |
| Delivery | A routing tool still needs real setup before a gate is trustworthy | An agency build usually takes several weeks, set per project | Ours has a fixed delivery time shown before you buy, with faster options |
| Coverage | A generic tool does not draft the action itself, only the routing | An agency deliverable depends on what was agreed | Ours has the AI draft the action and route it for approval |
Questions about this product
How much does Human Approval Workflow cost?
The build is one fixed price, and you can see it beside each delivery speed in the panel on this page. It covers the full Scope Card: one approval workflow of up to twelve steps, one AI drafting step, up to three connected systems and the review queue your reviewers use. The automation platform and AI usage that run it afterwards are billed to your own accounts.
When exactly does a human get pulled into Human Approval Workflow?
Every single time. This build is a standing gate, not a conditional rule, so every AI-proposed action stops at the review queue before it takes effect, and a named reviewer approves, edits or rejects it. If you only want a person pulled in for specific triggered situations, AI Escalation System is built for that instead.
What does Human Approval Workflow cost to run each month?
Running it afterwards costs roughly $20 to $100 a month in your own accounts, covering the automation platform plus AI model usage for drafting each proposed action. There is no separate platform fee on top of that figure. The exact cost depends on how many actions the AI proposes each month.
What access does Human Approval Workflow need to my systems?
We need admin or API access to the AI tool and every system involved in the workflow. Access is shared by adding our team as a user or creating a key you control, never by email. We also need a short description of how the AI tool's actions reach your systems today. Five to ten real proposed actions help us build a review queue that reflects genuine cases, not guesses.
Can we change the approval rules or reviewers later?
Yes. Adjusting who reviews an action, or refining what the AI is allowed to draft alone, usually fits inside the monthly change hours of maintenance. Adding an entirely new workflow, or a fourth connected system beyond the three in scope, goes through a Change Quote agreed in writing before that work starts.
Should I buy this or the Approval Workflow Layer add-on?
Buy this one when the AI came from another vendor or your own team. We build the approval gate and its connections from scratch around that tool. If the AI is an Upmark Labs product, choose the Approval Workflow Layer add-on instead. It adds the same approve, edit and reject step to the product we built, so nothing is rebuilt.
Who owns the approval history Human Approval Workflow records?
You do. Every approve, edit or reject decision is logged with who made it and when, inside your own connected systems, not a platform we control. We hand over an admin guide and a recorded session covering how the gate and routing work, and the full decision history stays visible to your team.
Guides on this topic
All guides Put a person in front of every AI action.
Order Human Approval Workflow and a standing approval gate in front of your AI process is live on a fixed delivery date.
Works well with
Oversight
Agent Exception Handler
Catches technical failures in AI runs, such as errors, timeouts and unusual runs, and routes them for resolution.