You own them only if two things are true: your contract assigns the rights in what the consultant builds to you, and the workflows run in accounts you control. Without a written assignment, the consultant often keeps the copyright and you get, at best, a licence to use the work. Ask for both before any build starts.
Those are two separate questions. Legal ownership decides who may copy, change and reuse the prompts, configurations and documents. Practical control decides who can log in, switch things off or move them. You can own the rights and still be locked out, or hold every login but only have a licence to the consultant's reusable templates. None of this is legal advice: have a solicitor check the final wording.
The parts of an AI workflow, and who tends to own each
"The workflow" is really several things with different owners by default. The middle column describes a common starting position in many legal systems when the contract says nothing; yours may differ, which is exactly why the contract should say something.
| Part | What it is | Common position if the contract is silent | What to ask for |
|---|---|---|---|
| Automation design | The Zap, Make scenario, n8n workflow or Power Automate flow: its steps, filters and settings | The running copy sits with whoever holds the platform account; rights in the design and documentation may stay with the consultant who created them | Assignment, built in your account, plus an exported copy |
| Prompts and instructions | The text that tells the AI what to do and how | Written by the consultant, so often theirs | Assignment, delivered as plain text files |
| Knowledge files | Your price lists, policies, FAQs and templates | Yours: you supplied them | Keep the originals; don't let them live only inside a tool |
| AI outputs | Drafts, summaries and replies the model produces | OpenAI's and Anthropic's terms assign you their rights in outputs, if any | Nothing extra from the AI vendor; the consultant shouldn't claim them |
| Accounts, keys and billing | Logins for Zapier, Make, OpenAI and similar | Whoever opened the account controls it | Open them yourself; add the consultant as a user |
| Documentation | Runbooks, diagrams, test sheets | Written by the consultant, so often theirs | Assignment, delivered in editable formats |
| The consultant's own library | Templates and components they reuse across clients | Theirs, and reasonably so | A permanent licence that survives the contract and a sale of your business |
In many legal systems, copyright in work made by an independent contractor stays with the contractor unless a written agreement transfers it, while work an employee creates as part of their job usually belongs to the employer. That default is why consulting contracts need an ownership clause, and why "we paid for it, so it's ours" isn't always true. For the related question of the content your business publishes, see whether you own AI-generated content.
Assignment or licence: what the difference means for you
An assignment transfers ownership: after it, the rights are yours, and the consultant needs your permission to reuse that specific work. A licence is permission to use something the consultant still owns, on the terms written down. Licences come in very different strengths, so the adjectives matter:
- Perpetual or for a term? A licence "during the engagement" can end the day you stop paying.
- Irrevocable or revocable? A revocable licence can be withdrawn, sometimes for reasons you don't control.
- Exclusive or non-exclusive? Non-exclusive is normal for a consultant's library: they reuse it with other clients.
- Transferable? If you sell the business, can the buyer keep using it? If not, the licence may disappear at exactly the moment its value is being counted.
- Can you modify it? Your next consultant will need to change the prompts and steps.
A practical threshold: anything built specifically for your business, and anything your operations depend on, should be assigned to you. A consultant's generic, reusable components can reasonably stay theirs, provided your licence is perpetual, irrevocable, transferable with the business and allows changes. A consultant who resists assigning your bespoke prompts and documentation, rather than their general library, is worth a careful conversation.
What decides how much this matters for your business
Ownership terms deserve more attention in some situations than others. Five factors, with rough thresholds:
- How central the workflow is. If it touches customers or money (enquiry replies, bookings, invoices, refunds), insist on assignment and accounts in your name. A workflow that tidies an internal spreadsheet can live with looser terms.
- How likely you are to change consultant. If there's any real chance, and over a few years there usually is, you need exported copies and the right to modify, or the next person starts from nothing.
- Whether you might sell the business. If a sale within a few years is possible, licences must be transferable, and a buyer's adviser will ask who owns the systems.
- How much of it is bespoke. A Zap with two standard steps is easy to rebuild; forty tested prompts encoding your policies are not. The more bespoke the work, the more the assignment matters.
- What data flows through it. If customer or health details pass through the workflow, you need to control the accounts for data-protection reasons as well as ownership ones.
Illustrative clause wording to take to your solicitor
These examples show what the clauses are trying to achieve. They are not legal advice, and wording that works in one legal system may not work in another, so have them adapted before use.
1. Ownership of deliverables
All rights in the Deliverables (including automation configurations,
prompts, instructions, test sets, documentation and diagrams created
for the Client) are assigned to the Client on payment of the
relevant invoice. The Consultant will sign any document reasonably
needed to confirm this.
2. Consultant's pre-existing materials
Where Deliverables include the Consultant's pre-existing templates
or components, the Consultant grants the Client a perpetual,
irrevocable, non-exclusive, royalty-free licence to use, copy and
modify them as part of the Deliverables, which survives termination
and transfers with the Client's business.
3. Accounts and credentials
All platform, automation and AI accounts used for the Deliverables
will be opened in the Client's name and billed to the Client. The
Consultant will access them only through user accounts granted by
the Client and will not hold the only copy of any credential.
4. Handover
Before final payment, the Consultant will provide exported copies of
all workflows, the final version of every prompt, a runbook for each
automation and a list of every account and connection used.It helps to recognise the weak version too. A clause such as "The Client may use the automations provided for the duration of this agreement" gives you a licence that ends with the contract, says nothing about changing the automations and doesn't mention prompts or documents at all. Nothing in it is unusual or sinister; it's simply written from the consultant's side, and asking to mark it up towards clauses 1 and 2 is a normal request.
The wider set of contract terms, including payment, liability and change requests, is covered in the nine AI consulting contract clauses to check, and agencies have their own version of the problem in AI clauses for agency contracts.
Whose account it runs in: the practical half of ownership
A perfect assignment clause won't help at 8am on a Monday if the automation runs in an account you can't log into. Each platform handles transfers differently, and the details decide how painful a move would be.
- Zapier. The Free and Professional plans allow one user; Team allows 25 with shared app connections. Within an account, "Change owner" moves a Zap to another user, and when a user is removed their Zaps and folders pass to an owner you designate or to the account owner. Between accounts, Zapier can copy Zaps, Tables, Forms, Interfaces, Chatbots and Agents, but connections arrive as placeholders to re-authorise, copied Zaps and agents arrive switched off, run history stays behind and any webhook URLs change, so anything sending data to the old address needs updating.
- Make. A scenario can be exported as a blueprint file containing its modules, settings and mapped values, and imported elsewhere, but blueprints don't include connections or credentials, which must be recreated.
- n8n. Workflows export as JSON files. The export includes credential names and IDs but not the secrets themselves; HTTP Request nodes imported from cURL commands can carry authentication headers, which should be removed before a file is shared.
- Power Automate. Flows belong to their owners. If an owner leaves, a flow can become "orphaned" and fail when its connections were tied to that person's account; admins can add co-owners, which is worth doing before anyone leaves.
- AI model accounts. Open the OpenAI or Anthropic developer account under a business email, pay with the business card, and add the consultant as a member of the organisation. API keys then belong to projects you control, and removing the consultant doesn't stop anything running.
- Custom GPTs. OpenAI is retiring them, with a hard stop on 11 December 2026. If a consultant built you one in their own workspace, it disappears with their access either way. Projects in your own ChatGPT or Claude workspace, Gemini Gems (becoming "skills" from November 2026, recreated automatically; sharing a Gem lets recipients view its instructions and uploaded files), or OpenAI's plugins and skills are the current routes for shared assistants.
The same checks apply to automations already running in your business. An audit of the Zaps and scenarios nobody owns often turns up at least one built under someone else's login.
Clean, messy and reasonable: three illustrative set-ups
These are illustrations to show how the pieces fit together, not accounts of real clients.
A personal trainer: mostly clean, one loose end. The consultant built a weekly check-in sequence in the trainer's own Zapier Professional account, kept every prompt in a document in the trainer's drive, and the contract assigned all deliverables. The one gap: the OpenAI API key used by the AI step sat in the consultant's developer account "for testing" and was never moved. It worked for five months, until the consultant tidied up old keys and the check-ins stopped. Fix: a new key in an organisation the trainer owns, the old key revoked, and a line in the handover list for next time.
A tattoo studio: messy. An agency ran all the studio's automations in its own Make organisation and charged a monthly platform fee. The contract gave the studio "a licence to use the automations during the term". When the studio moved on, the agency supplied blueprints as a courtesy, but no connections, and the licence had ended, so the studio's rights even to the enquiry prompts it had helped write were unclear. Rebuilding connections and testing took the studio's new helper a full day, and the prompts were rewritten from scratch to be safe.
A nail salon: reasonable. The consultant used a booking-reminder template from their own library and wrote salon-specific prompts on top. The contract assigned the salon-specific work and gave the salon a perpetual, irrevocable, transferable licence to the template. Everything ran in the salon's Zapier Team account with the consultant as a member, removed on the last day. Nobody owns everything, but nobody can switch anything off on the salon either.
A 15-minute check of what's already running
The trainer's loose end would have surfaced in a quarter of an hour, along with a second one nobody had noticed. For each automation, note three owners: the account it runs in, each app connection and any AI key. The trainer's check, done properly, looks like this:
| Automation | Account owner | Connections authorised by | AI key belongs to | Action |
|---|---|---|---|---|
| Weekly check-in messages | Trainer's business email | Trainer (Gmail, booking app) | Consultant's developer account | New key in the trainer's organisation; revoke the old one |
| New-client welcome email | Trainer | Trainer | No AI step | None |
| Payment reminder | Trainer | Consultant's test Gmail | No AI step | Reconnect with the business inbox |
Two of the three rows need action, and neither would have shown itself until something broke. A bigger business takes longer, but the same three columns apply.
Prompts, knowledge files and chat history
In AI workflows, the prompts are the closest thing to source code: they hold the rules, tone and exceptions you spent weeks agreeing. They deserve the same care as any other asset.
- Keep a master copy of every prompt as plain text in a folder you own, with a date and version number, even if the live copy sits inside a tool.
- If shared context lives in a ChatGPT or Claude Project, make sure it's in your business workspace, not the consultant's. On Claude Team, the Primary Owner can export the organisation's data, including members' chats, which is one reason to be the Primary Owner yourself.
- Don't count on exporting a ChatGPT Business workspace later: OpenAI says workspaces have no self-service data export. Anything important there should also exist as a file you own.
- Keep your own originals of every knowledge file. A price list that exists only inside an assistant is hard to update and harder to move.
- Store test sets alongside the prompts. When the next consultant changes a prompt, the tests show whether anything broke.
Supplier risk belongs in the same conversation. Tools do close: OpenAI discontinued Sora, and the calendar tool Clockwise shut down in March 2026 and deleted user data rather than transferring it. Owning exported copies of your workflows and prompts is what lets you move quickly if a platform goes, and what to do if an AI vendor shuts down covers the rest of that plan.
If the consultant won't assign the work
A refusal isn't automatically a bad sign. Consultants build their businesses on reusable components, and assigning those to every client would leave them nothing to work with. The real question is what exactly they want to keep. Reasonable positions to accept or propose:
- Split the work. Your business-specific prompts, test sets and documentation are assigned to you; their general templates and components are licensed.
- Strengthen the licence instead. If assigning some part is genuinely off the table, a perpetual, irrevocable, transferable licence with the right to modify gets you most of the practical benefit.
- Require exports regardless. Whatever the ownership terms, the contract can oblige them to hand over exported copies and the final prompts.
- Keep the accounts yours. This costs the consultant nothing and removes the worst risk: losing a working system the day the relationship ends.
What isn't reasonable is refusing all four. If a consultant insists on owning your bespoke prompts, keeping everything in their accounts and handing over nothing, the arrangement is closer to renting a system than commissioning one. That can be a legitimate product, but it should be described and priced as one, so you can compare it with the alternatives.
Four checks before you sign, and one at the end
- Who opens the accounts? You do, in the business's name, on the business's card. The consultant gets a user login.
- Is there an assignment clause for the deliverables? Prompts, configurations and documentation built for you should be assigned on payment.
- What are the licence terms for anything the consultant keeps? Perpetual, irrevocable, transferable with the business, and modifiable.
- Is the handover written into the contract? Exported copies, final prompts, a runbook and an account list, before the final payment.
And at the end: check you actually received it all. The AI consultant handover checklist turns that into a list you can tick off line by line on the last day.
Ownership questions owners ask after the build
Can a consultant reuse the workflow they built for me with other clients?
It depends on the contract. If the deliverables are assigned to you, the consultant can't reuse your specific workflow, prompts or documents, though they keep their general know-how and any pre-existing templates they only licensed to you. If there's no assignment, they may be free to reuse more than you'd like. Either way, a confidentiality clause should stop them reusing anything that reveals your data or business details.
Do I own what the AI produces inside my workflow?
As far as the AI vendors go, OpenAI's and Anthropic's terms assign you whatever rights they have in outputs, if any. Whether copyright exists in AI-generated material at all varies with the law where you are, so that part is uncertain. For day-to-day business drafts it rarely matters; for logos, designs or content you want to protect, ask a solicitor.
The consultant built everything in their own account and won't transfer it. What now?
Check the contract first: what it says about deliverables and handover decides your position. Then ask for exports: Zapier can copy Zaps to your account, Make exports blueprints and n8n exports workflow files, although connections must be rebuilt in each case. If the consultant refuses, get legal advice before rebuilding from scratch, and make accounts in your name a condition next time.
What if an employee built our automations, not a consultant?
In many legal systems, work an employee creates as part of their job belongs to the employer, so the rights question is usually simpler. The practical question isn't: automations built under an employee's personal login or email can stop when they leave. Move them into business-owned accounts while the person is still there, and add a second owner to anything that supports it.
Does signing an NDA settle who owns the work?
No. A non-disclosure agreement covers confidentiality: what each side may share or use from the other's information. Ownership of the workflows, prompts and documents needs its own clause, usually an assignment of the deliverables plus a licence for any pre-existing materials. Many consulting contracts contain both, but check that the ownership clause is actually there rather than assuming the NDA covers it.
Further reads
- Should You Sign an NDA Before Sharing Data With an AI Consultant? — The confidentiality side, which an ownership clause doesn't cover.
- Who Owns AI Chat History When an Employee Leaves? — The same ownership question for staff chat histories and prompts.
- Staff Offboarding Checklist for AI Tools and Shared Accounts — Removing access cleanly once ownership is sorted.
- How to Share AI Tool Logins Safely With a Password Manager — Sharing logins during the build without handing over control.
- AI Consultant Red Flags: 12 Warning Signs to Walk Away From — Building in the consultant's own accounts is one of twelve warning signs.
- What an AI Consultant Needs From You: Access, Data, and Time — How to give access that keeps you in control from day one.
- How to Evaluate an AI Implementation Proposal or Quote — A 22-point checklist for any AI implementation quote, with the phrases to pin down, a scoring sheet and two quotes compared over three years.
- How to Brief a Developer on a Custom AI Workflow You Need — The eight parts of a developer brief for an AI workflow, a copyable requirements template, and a worked example from a dog-grooming salon.
- Who Should Own AI in a Small Business? Roles and Responsibilities — The four roles AI needs in any small firm, how they're split in a barber shop, an optician and a garden centre, and a responsibilities chart to copy.
- Questions to Ask Before You Pay Anyone to Set Up AI Marketing — 30 questions in six groups, with what a good answer sounds like, plus a filled-in comparison of two real-looking quotes for a food truck.
- Should You Hire an Automation Consultant or Build It Yourself? — A five-question test, what DIY really costs in hours, one build that worked and one that didn't, and the hybrid route most small firms should take.
- Too Many Zaps? Signs Your Automations Need Consolidating — A 13-point automation sprawl checklist with a clinic's filled-in audit: 23 Zaps cut to 10 and monthly tasks more than halved, without losing a step.
- Power Apps for Small Businesses: When It Replaces Custom Software — When Power Apps can stand in for custom software: what Microsoft 365 already covers, where Premium licences start, and a vet practice's ward app costed.
- How to Choose an AI Consultant: 20 Questions to Ask First — Twenty questions to put to any AI consultant, what strong and weak answers sound like, and a scoring sheet filled in for a farm shop.
- Done-for-You vs Done-With-You AI Implementation: Which Suits You? — What done-for-you and done-with-you AI implementation leave you holding, how upkeep decides it, and a four-shop butcher that uses both.
- How to Scope an AI Project: Deliverables and Acceptance Criteria — A step-by-step method for scoping an AI project, with a furniture maker's one-page scope, acceptance criteria with pass marks and a testing process.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: OpenAI and Anthropic terms on output ownership; Zapier help on copying assets to another account and changing owners, and Zapier pricing (users per plan); Make help on scenario blueprints; n8n documentation on exporting workflows; Microsoft Learn on orphaned Power Automate flows; OpenAI's help pages on custom GPT retirement and ChatGPT Business data export; Google's help on sharing Gems and on the transition from Gems to skills.