Managed AI Services and ChatGPT Administration
An AI workspace is a system like any other: people join and leave, apps get switched on, a custom GPT quietly stops matching the process it was built for, and the seat bill grows while usage does not. Managed AI services give that work an owner. NetSys administers ChatGPT Business (formerly ChatGPT Team) and ChatGPT Enterprise workspaces, Microsoft 365 Copilot and Claude Team and Enterprise organizations under one month-to-month agreement, with a monthly report an owner can read.
The short answer
Managed AI services are the ongoing administration and support of a company's AI tools after rollout: user and seat management, policy enforcement, connected app and plugin reviews, usage and cost reporting, maintenance of custom GPTs and agents, and a help desk for staff questions. NetSys provides this for ChatGPT, Microsoft 365 Copilot and Claude workspaces under a month-to-month agreement, with a named owner on your side, a monthly report and help desk support seven days a week from 4 a.m. to 11 p.m. Eastern.
Closest related page: AI services for business. That page is the hub for everything NetSys does with AI, from consulting to custom builds; this page is the monthly service that runs and supports those tools after they are live.
Companies that have rolled out ChatGPT, Copilot or Claude and have nobody whose job is to run it. The owner or office manager who became the workspace admin by default. Firms with a custom GPT or an agent that someone built last year and nobody maintains. Managed IT clients who want their AI tools covered the way their network and Microsoft 365 tenant already are.
The service separates two kinds of work. Workspace administration is routine: seats, roles, offboarding, settings, reports. Maintaining GPTs and agents is engineering: testing them against a representative set of tasks, fixing instructions and integrations when a system changes, and reviewing the logs. Both are in the agreement, with the hours and inclusions written down, and new builds are quoted as projects.
Onboard, baseline, then run on a monthly cycle
Onboarding records who owns each workspace, which seats are assigned, which apps, plugins and custom tools exist and what the policy says, then fixes the obvious gaps: former staff still holding seats, apps nobody approved, GPTs with no owner. From there the service runs on a monthly cycle: joiners and leavers handled as tickets, settings checked against the agreed baseline, usage and spend read against the seat list, each GPT or agent tested and adjusted, and a short report sent to the named owner on your side with the decisions that need them. Changes outside the baseline wait for that owner's approval.
What managed AI services include
Users, seats and roles
The join and leave process, applied to AI.
- Seats assigned as part of the starter ticket, with the right role and seat type and the approved Projects, GPTs, agents or Skills shared to them
- Offboarding in the same ticket as the Microsoft 365 or Google account: seat removed, ownership of GPTs, Projects and integrations transferred, shared links reviewed
- Owner and admin roles kept to named people, with a quarterly check of who holds them
- The seat count reconciled against payroll and the vendor invoice each month, so you stop paying for people who left
Policy enforcement
The usage policy applied in settings, not only in a document.
- Workspace settings kept to the agreed baseline: training-data controls, sharing defaults, who may publish GPTs, agents or Skills, which apps are allowed, and the seat-type and feature permissions the plan offers
- Sign-in tied to company identity (single sign-on or domain verification where the plan supports it), so access ends when the account does
- Policy changes recorded with the reason and the approver
- Exceptions handled as tickets, with the named owner approving
Connected app and plugin reviews
Every connection re-checked after it is approved.
- A register of enabled apps, custom MCP servers, plugins, GPT Actions and agents, with scopes, owners and the date each was approved
- New app and plugin requests reviewed for what they can read and write before they are enabled, with the approval setting (always ask, read actions, low-risk actions or all actions), any per-app override and the policy for new actions recorded
- Domain restrictions kept current, so only company-domain accounts connect apps from the workspace where the plan supports it
- A scheduled review of the register, removing what is unused and re-checking scopes after vendor changes
- Microsoft 365 sharing re-checked before a Copilot agent or a SharePoint app reaches a wider group
Usage and cost reporting
Seats, usage and API spend on one monthly page.
- Active users per tool from the admin console's usage reports, set against the seats being paid for
- Credit and API spend for agents and integrations tracked against the estimate in their design, with caps where the platform allows
- Seat type recommendations (Standard or Premium on ChatGPT Business) when usage shows a license is wrong for the person
- Vendor price, plan and retirement notices flagged with what they mean for your bill and your tools
Monthly maintenance of GPTs and agents
Custom tools drift; maintenance catches it.
- Each custom GPT, workspace agent, Copilot agent and Claude-based agent run against its representative test set every month, with failures fixed
- Instructions and knowledge files updated when a process, price list or template changes
- Integrations re-tested after the target system or the vendor changes, including custom MCP servers that need an admin to refresh and republish
- Logs reviewed for wrong answers, blocked actions and unexpected cost, with findings in the monthly report
Help desk and incident response
A number staff can call when the tool does something odd.
- Staff questions about the AI tools go to the NetSys help desk, staffed seven days a week from 4 a.m. to 11 p.m. Eastern
- A suspected data leak, a compromised account or an agent acting wrongly is handled under the same incident process as the rest of your IT, with the agent or app paused first
- Vendor support cases opened and worked for you
- Refresher training scheduled when tools change or a team joins
Why businesses hand their AI workspace to NetSys
Tell us which AI tools are live, how many seats you pay for, which custom GPTs or agents exist and who administers the workspace today. A NetSys engineer will tell you what onboarding would check first and what the monthly service would cover.
Who does the work: meet the NetSys team on our About page.
- We stayed on after our published Copilot rollout to adjust the agents as the team used them and keep the tenant in step, which is the service this page describes
- One agreement covers ChatGPT, Microsoft 365 Copilot and Claude, with one register across all three
- The same engineers run your Microsoft 365 tenant, identity and security, so offboarding and app reviews happen in one process
- Inclusions, hours and exclusions written down, with new builds quoted separately so the monthly fee never hides a project
- A monthly report written for an owner, with the decisions that need them listed first
- Month to month, from one office in Brooklyn, in business since 1998
Managed AI Services in practice
- Importers and distributors
Regional food importing company
Microsoft Copilot and Claude rollout for 25 employees, AI usage policies, connector access controls, and staff training.
Microsoft Copilot ConsultingIT for Importers and Distributors
Client names are withheld. Each card is the scope of a real NetSys engagement, as delivered.
What is included each month, and what is scoped separately
The monthly service covers running the AI tools you have. Work with a start and an end, and anything the vendor charges for, is scoped separately so it can be planned and approved:
| Area | Included in the monthly service | Scoped separately |
|---|---|---|
| Workspace administration | Joiners, leavers, roles, seat reconciliation and settings kept to the agreed baseline | Setting up a new workspace or moving to another plan |
| Policy | Settings enforced, exceptions ticketed, changes recorded | Writing the governance program itself |
| Apps and plugins | The register maintained, requests reviewed, scheduled re-checks | New integrations, custom MCP servers and GPT Actions, quoted as projects |
| GPTs and agents | Monthly tests, instruction and knowledge updates, re-tests after vendor changes, and the list of what must migrate before OpenAI retires custom GPTs | New GPTs, agents or applications, and the migration or rebuild of each existing GPT, quoted per item |
| Reporting | A monthly usage, seat and spend report with the decisions listed | Custom dashboards beyond the report |
| Support | Help desk for staff questions, vendor cases, incident handling with the agent paused first | Training programs for new departments, scoped per cohort |
| Licenses and usage | Advice and reconciliation | The vendor's seat fees, credits and API usage, billed to you by the vendor |
The work on the right-hand side: ChatGPT consulting and business workspace setup · AI governance consulting and acceptable use policies · OpenAI and ChatGPT integration services · Custom GPT and workspace agent development · Corporate AI and ChatGPT training for employees.
For managed IT clients the service is added to the flat monthly fee per user; on its own it is quoted after a scoping call based on the tools, seats and custom tools maintained.
Onboarding and offboarding in an AI workspace
The same ticket that creates or disables a Microsoft 365 account triggers these steps, so neither waits on the other:
- Starter: a seat assigned with the right role and seat type, company sign-in enforced, the approved Projects, GPTs, agents or Skills shared, and the safe-use briefing scheduled.
- Role change: seat type and shared tools adjusted, and admin rights reviewed if the person held them.
- Leaver, first steps: the seat removed, owner or admin roles reassigned, and ownership of GPTs, Projects, Skills and integration credentials transferred to a named person.
- Leaver, follow-up: shared links and published tools the person owned reviewed, personal API keys and app authorizations revoked, and the paid seat count reduced at the next billing cycle.
- Record: every step in the ticket, so an auditor or insurer can see who had access to which AI tool and when.
What the monthly report covers
One page per workspace, written for the named owner, with the decisions first:
- Seats paid for versus active users per tool, with recommended changes
- Joiners, leavers and role changes handled, and any access still pending
- The app and plugin register: additions, removals and scopes changed
- GPT and agent test results, fixes made and anything still failing
- Credit and API spend against estimate, and any cap reached
- Incidents, vendor cases and policy exceptions, with their outcome
- Decisions that need the owner: new requests, renewals, retirements and vendor changes
Who owns what under managed AI services
A small firm often has one person as workspace owner and tool owner. The point is that each role has a name against it:
| Role | Held by | Accountable for |
|---|---|---|
| Workspace owner | A named person at your company | Approving seats, apps, new tools and policy exceptions |
| Tool owner | The manager of the team using a GPT, agent or Skill | Confirming it still matches the process and supplying updated documents |
| Administrator | NetSys engineers | Settings, seats, reviews, tests, reports and vendor cases |
| Help desk | NetSys | Staff questions, incidents and escalation, seven days a week |
| Counsel | Your in-house or outside counsel | Legal calls, such as retention obligations or client confidentiality |
AI consulting vs managed AI services
Many clients move from one to the other: the plan names the tools, the rollout sets them up, and the managed service keeps them running:
| AI consulting | Managed AI services | |
|---|---|---|
| What you get | A written plan: use cases, tools, fixes, pilot design | A running service: administration, reviews, maintenance, reports, support |
| When | Before a rollout, or when the next step is unclear | After the tools are live, for as long as they are used |
| Billing | A fixed-scope engagement | Month to month |
| Who does the work | Engineers who review and recommend | Engineers who administer and maintain |
| Typical trigger | Staff experimenting, licenses bought without a plan | Seats nobody reconciles, a GPT nobody maintains, an admin who left |
The consulting side: AI consulting services for small and mid-sized businesses.
What moves the monthly quote
NetSys publishes no price list. Vendor seat fees, credits and API usage are the vendor's charges and appear as their own lines. These are the inputs that move the monthly number:
| Factor | Why it moves the price |
|---|---|
| Tools and workspaces | ChatGPT, Copilot and Claude each have their own console, settings and reports |
| Seats and turnover | Joiners and leavers are the most frequent tickets |
| Custom GPTs and agents maintained | Each needs a test set, a monthly run and an owner conversation |
| Apps and integrations | Each one is reviewed on a schedule and re-tested after changes |
| Compliance evidence | HIPAA, SEC or insurer reporting adds scheduled reviews and records |
| Existing managed IT agreement | For managed clients the service is added to the per-user fee |
Paying for AI seats nobody reconciles?
Send the tools in use and the seat counts. We will tell you what onboarding would check first and what the monthly service would cover.
Managed AI Services FAQs
What are managed AI services?
Managed AI services are the ongoing administration, maintenance and support of a company's AI tools after they are rolled out, delivered by an outside team under a monthly agreement. At NetSys that means running your ChatGPT, Microsoft 365 Copilot or Claude workspace (seats, roles, settings, apps and plugins), maintaining the custom GPTs and agents built on them, reporting usage and spend each month, and taking staff questions at the help desk. An AI managed service provider is a company that does this; NetSys is a managed IT and cybersecurity provider, founded in 1998, that added it to its agreements because clients needed the same join, leave, review and report routine for AI that they already had for Microsoft 365. New builds and new workspaces are quoted separately.
What does ChatGPT administration involve?
On ChatGPT Business, administration happens in the workspace settings inside ChatGPT: Members (invite by email or CSV with role and seat type; owner, admin, analytics viewer and member roles), Billing (seats, invoices, credits, auto-reload, usage alerts, default seat type), Identity & access (domain verification, SAML or OIDC single sign-on, automatic account creation, workspace discovery), GPTs, apps and plugins (which apps are enabled, the approval setting per app from always ask to all actions, per-app overrides and the policy for new actions), Standard and Premium seat types, and Usage. Enterprise moves this into the Admin Console with groups, custom roles, SCIM, sign-in policies and the Compliance Platform. Two points shape the monthly routine: Business has no SCIM, so joiners and leavers are manual, and Business members can invite other members, which cannot be restricted, so the member list needs a regular review. Source: OpenAI help center, managing members, seat types and roles in ChatGPT Business, read October 17, 2026.
How do you offboard an employee from ChatGPT or Claude?
In the same ticket that disables their Microsoft 365 or Google account. In ChatGPT Business we remove the member, which frees the seat but does not reduce the paid seat count, so we schedule the reduction for the next billing cycle (seats added on an annual plan stay committed through the term). OpenAI reassigns the person's Projects and GPTs to a workspace owner and retains their chats, files and canvas documents indefinitely, which the runbook records. On Enterprise, SCIM removes access when the directory account is disabled and the configured retention policy applies. In a Claude Team or Enterprise organization we remove the member and transfer ownership of shared Projects and integration authorizations. Each step goes in the ticket. Sources: OpenAI help center articles on billing and seats and on data retention when a member is removed, read October 17, 2026.
Is this the same as managed AI security or AI-powered detection products?
No. Some security vendors sell managed AI defense or AI-driven detection and response; that is endpoint and network security operated for you, which NetSys covers under managed detection and response. Managed AI services on this page are about running the AI tools your staff use. The two meet at offboarding, app reviews and incident handling, which run under one process for managed clients.
How is managed AI services pricing structured?
Month to month, like every NetSys agreement. For managed IT clients the service is added to the flat monthly fee per user; on its own it is quoted after a scoping call based on the tools and workspaces covered, the number of seats, how many custom GPTs and agents need monthly maintenance and any compliance reporting. Vendor seat fees, credits and API usage are billed to you by the vendor; on ChatGPT Business, for example, credits bought for advanced features are valid for 12 months and spend limits per seat type are unset until an admin sets them. NetSys publishes no price list. Source: OpenAI help center, flexible pricing and spend controls, read October 17, 2026.
Do you support Copilot and Claude under the same agreement as ChatGPT?
Yes. The register, the monthly report and the offboarding process cover all three, and most clients run at least two: a chat workspace plus whatever is built into Microsoft 365. Copilot administration includes the tenant-side work (SharePoint sharing, sensitivity labels, agent publishing), because Copilot surfaces only what each user already has permission to view. Claude administration covers the organization's owners, connectors, Projects and provisioned Skills; Claude Team is 2 to 150 seats with single sign-on and domain capture, and SCIM, audit logs and custom retention are Enterprise features. Sources: Microsoft Learn on Copilot data and privacy, and Anthropic's Team and Enterprise plan articles, read October 17, 2026.
What are the support hours?
The NetSys help desk is staffed seven days a week from 4 a.m. to 11 p.m. Eastern, and monitoring and emergency service run 24/7. An agent acting wrongly or a suspected data leak is treated as an incident at any hour: the agent, app or key is paused first, then the review happens. Routine requests such as a new seat or an app review are handled during help desk hours.
Will you maintain a custom GPT or agent that someone else built?
Usually, after an onboarding review. We document what it does, who owns it, what it can reach and how it is tested, write a representative test set if none exists, and fix the gaps we find before it joins the monthly cycle. For custom GPTs there is a deadline: OpenAI has scheduled them to retire on December 11, 2026 (as of October 2026) in favor of plugins, so each one is either migrated, rebuilt as a workspace agent or retired, and the migration is quoted per GPT. A tool whose owner has left and whose purpose nobody can state is retired rather than maintained. Source: OpenAI help center article on the GPT retirement, read October 17, 2026.
Sources and technical references
- OpenAI help center: managing members, seat types and roles in ChatGPT Business
- OpenAI help center: managing billing and seats in ChatGPT Business (removing a member does not reduce the paid count)
- OpenAI help center: data retention when a member is removed from a workspace
- OpenAI help center: SCIM provisioning (not available on standalone ChatGPT Business)
- OpenAI help center: spend controls in ChatGPT Business
- OpenAI help center: flexible pricing and credits for Business, Enterprise and Edu
- OpenAI help center: the Compliance Platform for Enterprise and Edu
- OpenAI help center: custom MCP servers (the tool snapshot an admin must refresh)
- OpenAI help center: the scheduled retirement of custom GPTs and migration to plugins
- OpenAI help center: multi-factor authentication (enforced through the identity provider)
- Anthropic: the Claude Team plan (2 to 150 seats, SSO, domain capture)
- Anthropic: the Claude Enterprise plan (SCIM, audit logs, custom retention, Compliance API)
- Microsoft Learn: data, privacy and security for Microsoft 365 Copilot
Guides on this topic
- ChatGPT consulting and business workspace setup
- Corporate AI and ChatGPT training for employees
- Custom GPT and workspace agent development
- OpenAI and ChatGPT integration services
- AI workflow automation
- AI agent monitoring: quality, access, cost and monthly maintenance
- AI agents for business: consulting and development
- Microsoft 365 Copilot consulting and rollout
- Claude consulting and business implementation
- AI governance consulting and acceptable use policies
- Managed IT services: the agreement this can join
- The NetSys help desk
- Onboarding, offboarding and access reviews
- Managed detection and response (security monitoring)
- Case study: Copilot rollout with ongoing agent adjustment for 18 people
- AI services overview
Related services
Give your AI workspace an owner.
Tell us the tools in use, the seats you pay for and the custom GPTs or agents you rely on. We will say what onboarding would check first, what the monthly service covers and what would be scoped separately.