Buyer guide · platform decision
Exit cost, governance evidence and total cost decide this purchase more than feature lists do. Below: a rubric you score yourself, three decision paths, and a cost worksheet whose formulas you can read. ZeroTwo publishes this page and is one of the options it discusses.
Score each axis 1 to 9
Disclosure
ZeroTwo sells a multi-model workspace and publishes this page. We do not score other vendors. The rubric is yours to fill in, and where ZeroTwo does not fit we say so.
Figures read on each publisher's own page on 5 October 2026. They describe the market, not ZeroTwo.
Score each axis from 1 to 9 against the anchors, once per platform on your shortlist. Anchors, not adjectives: if you cannot find evidence for a 9, it is not a 9.
| Axis | Question | Strong (9) | Middling (5) | Weak (1) |
|---|---|---|---|---|
| 01Model concentration | How tied is the platform to one provider's roadmap? | Several providers are first-class and you can change the default model per task. | Two providers are wrapped; a third needs custom integration work. | One provider's models behind an open-sounding brand. |
| 02Stack-exit cost | If you leave in 18 months, what comes with you? | Prompts, projects, files and logs export in documented, standard formats. | Prompts export; configurations and tuned assets stay behind. | No export path; everything lives in the vendor's tenant. |
| 03Governance evidence | What can you verify before the pilot, not after? | Current third-party audit reports, a sub-processor list, a signed DPA, audit logs and role-based access, all provided on request. | A report exists but its scope does not clearly cover the product you will use. | A badge or a marketing page and nothing you can hand to your security team. |
| 04Multi-model parity | Can a non-engineer try several models on the same task? | Any model is one selection away inside the same surface and conversation flow. | A model picker exists, but each switch means a fresh setup. | One model only; trying another means buying another product. |
| 05Time to first outcome | How long from signed PO to a real user doing real work? | Days: identity and connectors are configuration, not a project. | Weeks: a partner-led implementation is expected. | Months: every connector is a paid engagement. |
| 06Price predictability | Seat, credit pool, token meter or negotiated agreement? | Published list price per seat and a clear rule for what happens at the allowance. | Seat plus a meter, with overage you can estimate. | Pure consumption with no ceiling by default. |
| 07Workforce reach | Does it serve the engineers and the other 90% of staff? | A chat-grade interface for everyone, with an API for builders. | Strong for one group, awkward for the other. | A developer sandbox presented as a company-wide platform. |
No scores. For each axis: what this site can point to, and what you should verify yourself before relying on it.
| Axis | What we can point to | What to verify |
|---|---|---|
| 01 Model concentration | 60+ models from several providers, including OpenAI, Anthropic, Google, xAI and Mistral, listed on the models page. | Which models your plan includes. Access widens by tier (Free is limited), so ask which are enabled for a Business workspace. |
| 02 Stack-exit cost | Nothing on this site documents export formats for chats, projects or prompts, so we make no portability claim. | Export formats, deletion timelines and what happens to files at termination, in writing, before you sign. |
| 03 Governance evidence | This page asserts no ZeroTwo certification. The Business page mentions SSO and multi-factor authentication and sends SCIM, role-based access and data-residency questions to sales. | Audit reports and their scope, the sub-processor list, DPA terms, retention windows and training-data terms. If you need FedRAMP, ZeroTwo does not claim it. |
| 04 Multi-model parity | 60+ models across providers are available in one workspace; plan limits decide which ones you can use. | Run the same ten real tasks on two or three models yourself. Do not rely on a vendor's ranking, including ours. |
| 05 Time to first outcome | The Free plan needs no card (100 free credits to start, limited models), so a pilot owner can try the product before a purchase order. | Whether the identity and connector setup you need is available on your plan. Measure days from PO to first production user in the pilot. |
| 06 Price predictability | Business lists at $39.99 per seat per month, $35.99 per seat per month with annual billing, with 4,000 monthly credits per seat pooled across the team. Purchased credits never expire. | What happens when the pooled credits run out, volume pricing, and the renewal escalator. |
| 07 Workforce reach | Web app, desktop apps for macOS (Apple Silicon only), Windows and Linux, plus iOS and Android. | Your device fleet. Intel Macs are not covered by the desktop build. |
Plan and price details: ZeroTwo pricing. Model list: ZeroTwo models.
Three buyer paths, each with the questions that decide it. Pick the path first and shortlist inside it. Product names are examples for orientation, not rankings.
Path 1
e.g. Azure AI and Copilot Studio, Google Vertex AI, AWS Bedrock
Start with the suite you already own. Keep a multi-model workspace as a comparison tool for teams that want to test models outside the suite.
Path 2
e.g. Salesforce Agentforce, IBM watsonx, Palantir AIP
Evaluate the platforms with the deepest hooks into your system of record. Depth of integration matters more here than model breadth.
Path 3
e.g. Single-vendor (ChatGPT Enterprise) or multi-model (ZeroTwo)
A workspace fits. If one provider is enough, a single-vendor workspace is a defensible choice. If you want model choice, shortlist multi-model options such as ZeroTwo.
Once the platform is chosen, the next question is which tools sit on top of it for each department. Work through that in map enterprise AI tools to your approved platform.
We do not publish totals for other vendors: the inputs depend on your quote, your usage and your team. What we publish is the structure, the formulas and the one set of numbers we can source, ZeroTwo Business list price.
$119,970
250 seats at the monthly-billing list price, per year
$107,970
250 seats at the annual-billing list price, per year
1,000,000
pooled credits a month across those seats
Illustrative headcount of 250. Seat licences only: implementation, the team that runs the platform and any extra credits are separate lines below.
| Cost line | Formula | Input comes from | ZeroTwo Business example |
|---|---|---|---|
| Seat licences | seats × price per seat per month × 12 | Quote or public list price | 250 × $39.99 × 12 = $119,970 ($107,970 at the annual-billing price) |
| Metered usage | seats × active share × prompts per month × tokens per prompt ÷ 1,000,000 × blended price per million tokens × 12 | Your own usage sample and the vendor's rate card | Not applicable to a seat with a credit pool: 1,000,000 pooled credits a month at 250 seats. Extra credits are bought separately; see the pricing page. |
| Overage or true-up | contract terms × expected overshoot | Contract draft | Ask what happens when the pool runs out and how it is billed. |
| Implementation | partner hours × blended hourly rate | Statement of work | No implementation fee is published. Ask sales for a written answer. |
| Platform team | FTE × loaded cost per FTE per year | Your own payroll assumptions | Your own number. It is the line that swings build-it-yourself options the most. |
| Security and procurement review | review hours × internal hourly cost | Your own history on similar vendors | Your own number. |
| Exit reserve | migration hours × rate | Answer to rubric axis 02 | Depends on the export formats you get in writing. |
The CSV has three scenarios side by side with live formulas. Scenario C is prefilled with ZeroTwo Business list price; fill A and B from your quotes. About one in five respondents in McKinsey's 2026 survey say AI operating costs have limited their use, so fill the metered-usage row with a pilot sample, not a guess.
Clear the evidence items before the technical evaluation; the cadence is a suggestion, not a benchmark.
Week 0
Security has signed off and identity is wired. Name a business owner for the first workflow, not a sponsor.
Week 2
One team, one daily task, a measurable before and after. No all-hands launch yet.
Week 6
Short, written training. Shadow-AI policy updated to point at the sanctioned tool.
Week 12
Review usage and the financial outcome attached to the two live workflows. Stop spend that cannot be tied to an outcome.
Our view, labelled as such.
An enterprise AI platform is a governed environment where an organization can run generative AI and agent workloads on one or more models, with identity, audit, data-handling and contract terms that procurement and security teams can verify. The test is evidence you can hand to your security team, not the word enterprise on a pricing page.
Three things change: verifiable governance (audit reports, sub-processor lists, a signed data processing agreement), identity and access tied to your corporate directory, and contract terms on training data, retention and exit. A consumer plan is a product one person buys; an enterprise platform is a contract an organization signs.
This guide uses three buyer paths. Hyperscaler suites such as Azure AI, Google Vertex AI and AWS Bedrock. Agent platforms such as Salesforce Agentforce, IBM watsonx and Palantir AIP. Company-wide workspaces, either single-vendor such as ChatGPT Enterprise or multi-model such as ZeroTwo. The names are examples for orientation, not rankings or endorsements.
There is no honest single number, because the total depends on seats, usage, implementation and the team you need to run it. Use the worksheet on this page and fill it from quotes. For reference, ZeroTwo Business lists at $39.99 per seat per month, so 250 seats cost $119,970 a year in seat licences at list price with monthly billing, before any implementation or platform-team cost.
It depends on whether you can tie it to an outcome. Forrester predicts enterprises will defer about 25% of planned AI spend into 2027 and reports that fewer than one-third of decision-makers can tie AI value to financial growth. A practical rule is to attach a financial outcome to each production workflow by the first business review and stop spending on the ones that cannot show one.
If you have one identity provider, no government authorization requirement and mostly knowledge-work use cases, a company-wide workspace is usually the shortest path to a first outcome. If the work lives inside Salesforce, ServiceNow or SAP, evaluate the agent platform tied to that system first.
Yes, that is what multi-model workspaces are for. ZeroTwo offers 60+ models from several providers in one product. Which models you can use depends on your plan, so check the pricing page and ask sales about a Business workspace.
Request the audit reports, sub-processor list and data processing agreement before the technical evaluation, and agree data residency, retention and training-data terms in writing. Those items cause most delays; once they are cleared, the rest is implementation.
Map each department to a tool type, then bring the shortlist back to your approved platform.
The multi-model workspace category in more depth.
How agents are built, supervised and run in production.
A wider look at platforms across consumer and business use.
Team plan, pooled credits and shared workspaces.
Side-by-side pages against individual AI products.