You are deciding how your organisation gets its first working AI agent: build it on Microsoft Copilot Studio, or start with a pre-configured AI worker such as UD's AI Staff. Here is what that decision actually turns on, with prices and limitations checked on 8 October 2026.
This is not a neutral question for a department head. Gartner's 2026 CIO and Technology Executive Survey found only 17% of organisations had actually deployed AI agents, against more than 60% expecting to within two years. The gap is rarely ambition. It is the time and skill needed to get from licence to a result you can report upward.
What is the difference between Copilot Studio and AI Staff?
Copilot Studio is a platform for building your own agents inside the Microsoft ecosystem, billed by usage. UD's AI Staff is a set of pre-configured AI roles that UD sets up for you in a sandbox. One gives you a toolkit and maximum control; the other gives you a working role quickly with less to build.
The distinction matters because the two products answer different buyer questions:
--- Copilot Studio answers: "We have Microsoft 365, a Power Platform team and specific processes. How do we build agents that fit them exactly?"
--- AI Staff answers: "We need one function, such as customer service or HR enquiries, handled by AI this quarter, without a development project."
If you are comparing Copilot Studio against other ecosystems rather than against managed services, our Microsoft Copilot alternatives guide covers that decision.
How much does Copilot Studio cost in 2026?
Copilot Studio is billed in Copilot Credits at US$0.01 each on pay-as-you-go, or US$200 per month for a 25,000-credit capacity pack. A typical knowledge response costs 2 credits, a response grounded in your Microsoft 365 data costs 12, and each tool call costs 5.
Copilot Studio moved to usage-based Copilot Credits on 3 August 2026, according to Anders Jensen's pricing analysis, which draws on Microsoft's Credits Guide. Key rates and terms:
--- Pay-as-you-go: US$0.01 per credit, billed monthly through Azure, no commitment.
--- Capacity pack: US$200 per month for 25,000 credits. Unused credits do not carry over.
--- Pre-Purchase Plan: annual prepayment with a 5% to 20% volume discount, starting at 300,000 credits.
--- Tenant graph grounding: 12 credits per response using SharePoint, Outlook or Teams data.
--- Other knowledge: 2 credits per response.
--- Tool or MCP call: 5 credits per invocation.
A worked example from the same analysis: 100 users holding 20 conversations each per month comes to roughly 15,000 credits, about US$150 (approximately HK$1,170) per month for one agent. The pack is listed in the FAQ on Microsoft's Copilot Studio pricing page rather than on a price card.
How much does AI Staff cost to start?
UD AI Staff has a published activation fee of HK$0, requires no credit card at registration, and puts one working AI role in a sandbox within three business days. Ongoing pricing after the trial is not published, so a buyer must ask for a written quote before committing to production.
The verified trial facts, described in detail in our AI Staff enterprise trial guide:
--- Activation fee: HK$0.
--- Credit card: not required at registration.
--- Time to a working role: three business days.
--- Roles available: Marketing, HR, Admin, Accounting, Customer Service and IT.
--- Languages: English, Traditional Chinese and Simplified Chinese.
--- Environment: a sandbox isolated from production data.
--- After the trial: a scoped commercial agreement or walk away; no automatic conversion to a paid plan.
The honest gap is the month-four number. HK$0 covers the start, not the run rate. Put the ongoing price question in your first call.
What does the first quarter cost under each option?
For a modest internal agent, Copilot Studio's credit bill is often small, sometimes under US$50 a month. The larger cost is the people who build, test and maintain it. AI Staff removes most build effort for the pilot, but its post-trial price is unknown until quoted.
Take an illustrative 150-person professional services firm automating HR policy enquiries. Assume each employee asks four questions a month, so 600 conversations.
--- Copilot Studio credits: if half the answers draw on SharePoint (12 credits) and half on uploaded files (2 credits), that is about 4,200 credits, roughly US$42 per month on pay-as-you-go.
--- Copilot Studio build effort: a maker designs topics, connects knowledge, tests and publishes. On the new experience, those test conversations are billed too.
--- Licence effect: if all 150 staff already hold Microsoft 365 Copilot licences, classic-experience internal usage is exempt from credits.
--- AI Staff: HK$0 to activate and three business days to a sandbox role; production pricing to be quoted.
Either way, budget two to four hours a week of a process owner's time to review output in the first month. That cost decides success more than the licence line.
Is Copilot Studio better than AI Staff? A side-by-side comparison
Neither is better in general. Copilot Studio is stronger on flexibility, Microsoft 365 integration and governance controls. AI Staff is stronger on speed, upfront cost and the build effort required. The right choice depends on whether you have builders, Microsoft 365 data and a clear process to automate.
Comparison on the same criteria
--- What you get: Copilot Studio, a build platform. AI Staff, a configured AI role.
--- Starting cost: Copilot Studio, pay-as-you-go from the first credit. AI Staff, HK$0 activation.
--- Ongoing cost visibility: Copilot Studio, published per-credit rates. AI Staff, quote-based, not published.
--- Time to first result: Copilot Studio, depends on your build team, typically weeks. AI Staff, three business days to a sandbox role.
--- Skills required: Copilot Studio, a maker or Power Platform team. AI Staff, a process owner and a reviewer.
--- Microsoft 365 grounding: Copilot Studio, native. AI Staff, not its design centre.
--- Local support: Copilot Studio, via Microsoft partners. AI Staff, UD's Hong Kong team configures it.
Where does Copilot Studio win?
Copilot Studio wins when your knowledge already lives in SharePoint and Teams, when you have people who can build, and when you need fine-grained control over what agents do. Its new hooks feature and native connectors give IT a level of control a packaged service will not match.
Be clear about these advantages before you choose:
--- Native grounding: agents answer from SharePoint, Outlook and Teams without a separate integration project.
--- Licence leverage: on the classic experience, internal users with a Microsoft 365 Copilot licence are exempt from credit billing.
--- Deterministic controls: Copilot Studio has added hooks in preview, which run a workflow on events such as session start, before a tool call and after a tool call, so IT can allow, modify, deny or audit actions.
--- Breadth: one platform can serve many departments, with connectors, MCP servers and agent flows.
For a 400-person bank or insurer with a mature Microsoft estate and a Power Platform centre of excellence, these points usually decide the matter.
What are the honest limitations of each option?
Copilot Studio's main limitations are cost predictability and build effort: credits are pooled, packs expire monthly and the new experience bills during building. AI Staff's main limitations are unpublished ongoing pricing, a single-role sandbox and no deep legacy integration during the trial.
Copilot Studio limitations
--- Credits sit in one tenant-wide pool, shared with other Copilot products, so one busy agent can consume another team's budget.
--- Capacity pack credits expire monthly; use half and your effective rate doubles to US$0.016 per credit.
--- On the new default experience, preview chats and evaluations are billed like real usage, and the licence exemption does not apply.
--- You still need people to design, test and maintain agents. The platform does not supply the process knowledge.
AI Staff limitations
--- Ongoing pricing is not published; you need a quote before production.
--- The trial covers one role in a sandbox, not a cross-department rollout.
--- Deep integration with legacy ERP or core banking systems is outside the trial scope.
--- A single-role pilot does not prove the vendor can run five departments at once. Ask about that before signing beyond the pilot.
Which should you choose? A verdict by buyer type
Choose Copilot Studio if you are Microsoft-centric with in-house builders and Microsoft 365 Copilot licences. Choose AI Staff if you need one function automated this quarter without a build team. If you are unsure, pilot a packaged role first to prove value, then decide whether to build.
--- Microsoft-heavy enterprise with a Power Platform team: Copilot Studio. Use the licence exemption and native grounding.
--- Department head with budget for a pilot, no developers: AI Staff. Three business days to a role you can measure.
--- Data mostly outside Microsoft 365: AI Staff or another non-Microsoft option. Copilot Studio starts one integration behind.
--- Regulated workflow touching core systems: neither alone. Scope integration and data governance first.
--- Undecided leadership: run a packaged pilot to create evidence, then evaluate building at scale.
Remember the base rate. Gartner has predicted that over 40% of agentic AI projects will be cancelled by the end of 2027, citing unclear business value and inadequate risk controls. The first agent's job is to produce a clear, reportable result.
Quick facts: Copilot Studio vs AI Staff (checked 8 October 2026)
--- Copilot Studio pay-as-you-go: US$0.01 per Copilot Credit.
--- Copilot Studio capacity pack: US$200 per month for 25,000 credits, no carry-over.
--- Copilot Studio typical rates: 2 credits per knowledge response, 12 per Microsoft 365 grounded response, 5 per tool call.
--- AI Staff activation: HK$0, no credit card.
--- AI Staff time to first role: three business days, in a sandbox.
--- AI Staff roles: Marketing, HR, Admin, Accounting, Customer Service, IT.
--- AI Staff ongoing price: not published; quote required.
Conclusion: pick the path that gets you evidence fastest
Copilot Studio is the right long-term platform for many Microsoft-centric enterprises. It is not always the fastest way to a first result. AI Staff is a fast, low-commitment way to prove one function, with the open question of ongoing price. The correct next step is to name one process, one owner and one success metric, then choose the option that can show that metric within a quarter.
We understand the cold edges of AI and the hard parts of your work, and UD has walked with Hong Kong enterprises for twenty-eight years, making technology a partnership with warmth.
Reviewed by the UD enterprise AI team. Copilot Studio rates were checked on 8 October 2026 against Microsoft's pricing page and the dated third-party analysis linked above; AI Staff terms against UD's published trial details. Vendors change pricing often, so confirm before buying.
Start a Free AI Staff Pilot
HK$0 activation, no credit card and a working AI role in three business days. We'll walk you through every step, from choosing the first role and setting success metrics to sandbox configuration and a written quote for production.
