Rolling out Digital Minds across an organisation

Roll out Digital Minds in four phases: pilot one Mind with named owners, set governance before you add anyone, expand in small groups, then review on a fixed schedule. In Plandora, one Enterprise contract can hold many Minds, each kept by its own account, with Owner and Member roles and control over who can open each Mind.

What are the phases of a rollout?

Treat each phase as a gate. You move on when its exit check passes, and the calendar has no say.

1. Pilot one Mind. Pick one workflow, one audience and one person accountable for its knowledge. Plandora works with larger teams through a guided rollout (see Enterprise) that begins with one valuable workflow, its knowledge owners and users, permissions and integrations, human review and escalation, and a measurable production outcome. The team validates the workflow and provider setup before usage expands. Keep the pilot Mind Private, so the public can't open it while you test. Private is not a pilot group, though: everyone in your organisation can still open it. Exit check: every handoff path has been tried in the chat preview, someone has taken over a conversation in Inbox, and the owner has read a sample of real conversations.
2. Set governance. Before anyone else is invited, write down who owns each topic's knowledge, who approves Playbook changes, what each department's Mind may answer, who staffs handoffs, how often you review and who answers for data protection. The checklist below covers each one. Exit check: every item has a named person and a date.
3. Expand in groups. Invite the next Minds in small waves. Each enrolled Mind takes one seat and pending invitations reserve seats, so size a wave by how many people can review knowledge and cover handoffs. Automations, integrations and extra channels may need setup and validation, depending on your plan or rollout, so add them to a Mind that already behaves well. Decide the human judgement first and automate it second; automation should follow judgment sets out why. Exit check: each new Mind has its own knowledge owner and handoff roster before it goes public.
4. Review. On the agreed day, read conversations, check the gaps Insights surfaces and change one thing at a time. Then record a decision for each Mind: expand, hold or unpublish. Exit check: a written decision from the accountable person.

How do organisations work in Plandora?

An organisation is a team's Minds under one contract. Each person keeps their own account and Mind, and the contract pays for the Minds enrolled in it. You request an organisation from the account menu, and nothing changes until the contract is activated. At the request you choose between an existing Mind and a separate organisation Mind, so decide early which one your pilot will use: the ownership transfer dialog shows what happens to the organisation's Mind when ownership changes hands.

Membership has an exit. A Member can leave, or an Owner can remove a Mind, and either way the Mind returns to its owner's personal plan, new work follows personal limits and a Private Mind is unpublished. Keep the approved source documents in your own systems, so the organisation's knowledge doesn't live only inside a Mind that may leave.

Plandora's enterprise FAQ says the scope can include single sign-on, role-based access, private connectors and a data-residency review. Treat each as something to confirm for your configuration, in writing, rather than assume. Bring identity, residency and procurement requirements to that conversation before sensitive or regulated material is uploaded.

What belongs on a governance checklist?

Seven items, each with a named person.

• Who owns the knowledge. One owner per topic area, such as pricing, policies and service descriptions, plus a review date. Facts live in Knowledge and behaviour lives in the Playbook, so the owner's job is to keep the facts current.
• Who approves Playbook changes. The Playbook has four sections (Goal, Role & style, Boundaries and Flow), each holding up to 4,000 characters, with 12,000 in total. That is short enough to read end to end at every change. The organisation guide describes Owner and Member roles for Minds, invitations, visibility, brand and billing. Approval of Playbook edits is yours to define: use your own change process and record who approved what and when.
• Boundaries by department. Where it fits, give each team or audience its own Mind with its own Boundaries: what it may answer, what it must decline and which topics always go to a person. Each Mind belongs to one account, so a Mind per team means an account per team. Plandora's own guidance is that a Digital Mind is not the human expert and should not be presented as one, and that a person should stay involved in financial, legal, medical and other consequential decisions.
• A handoff roster. Tag named people or teams in the Playbook so a handoff reaches someone specific. In Inbox, staff can take over, assign a conversation to a colleague and release it back to the AI, and the Unreplied filter shows who is waiting. Decide coverage hours and what visitors are told outside them. Human handoff for AI conversations goes deeper.
• A review cadence. A fixed day, named attendees and the same agenda each time: questions with no or partial answers, handoffs, knowledge changes. Insights offers 7-day, 30-day and 90-day periods and compares each with the previous period of the same length.
• A data protection owner. One person answers for what is uploaded, who can see conversations, visitor notices, retention and data-rights requests. Plandora's FAQ says authorised owners and organisation members may be able to review conversations and audience records according to their permissions, and that the deploying business is responsible for notices, permissions and lawful use of collected data. Rules differ by country, so start with data protection for AI chat in Southeast Asia and ask qualified counsel about your own case.
• Succession and offboarding. Decide who takes over when a Mind's owner leaves, and require an Owner to transfer ownership before they go.

What do NIST and ISO add?

They name the pieces so you don't have to invent them. Neither is a Plandora feature, and nothing here claims that Plandora or any rollout is compliant or certified.

The NIST AI Risk Management Framework (opens in a new tab)^[1], released on 26 January 2023, is intended for voluntary use, and NIST's page says version 1.0 is being revised, so check which version your reviewers work from. Its Govern function (opens in a new tab)^[2] is described as "a cross-cutting function that is infused throughout AI risk management", which matches the order above: governance comes before the pilot and continues through every phase. Three subcategories map onto this guide:

• Govern 2.1 asks that roles, responsibilities and lines of communication be "documented and are clear to individuals and teams throughout the organization". The RACI table below is one way to do that.
• Govern 1.5 asks that periodic review be planned, "including determining the frequency of periodic review". That is your review cadence.
• Govern 1.7 covers decommissioning and phasing out AI systems safely, which is the unpublish decision in phase four.

NIST also notes that "Senior leadership sets the tone for risk management within an organization", which is the sponsor's job in the table below. It adds that its actions "do not constitute a checklist, nor are they necessarily an ordered set of steps", so use it to check coverage rather than to tick boxes. For generative AI specifically, NIST published a Generative AI Profile (opens in a new tab)^[3] on 26 July 2024 as a companion resource to the framework.

ISO/IEC 42001:2023 (opens in a new tab)^[4], published in December 2023, specifies requirements for establishing, implementing, maintaining and continually improving an AI management system. ISO calls it "the world’s first AI management system standard" and says it suits organisations of any size that develop, provide or use AI-based products or services. It is a management-system standard, so it applies to how your organisation runs AI rather than to any one product. Whether to adopt it is a decision for your risk, compliance and procurement teams.

Who is responsible for what?

Plandora's roles are Owner and Member. Every other column below is a duty you assign, and one person can hold several. The cells follow the RACI convention in plain words: Lead is accountable for the outcome (one per row), Do is responsible for the work, Ask is consulted and Tell is kept informed. The Organisation Owner column assumes your Plandora Owner sits in IT or operations.

What should you measure?

Set the measures before the pilot starts, and record a baseline from your current process, such as how long enquiries wait for a human reply today. These are things to measure, not results to expect.

Read a sample of conversations at every review, whatever the numbers say. Counts point you to a question, and the answer is in the conversation.