Business software · Seats, usage and the person operating it
16 Best AI Tools for Businesses (2026)
Which AI tools should your business buy?
The best AI tools for businesses usually start with your existing stack: Copilot for Microsoft 365, Gemini for Google Workspace, or ChatGPT Business or Claude Team for a separate assistant. Buy automation only with a named operator. Susan is our first pick when you want a recurring job run for you rather than software your staff must operate.
Compare the buying unit before the price. The shortlist below uses public prices checked November 16, 2026; annual equivalents require the stated commitment. Each option links to its full pricing constraints and primary sources. USD and EUR remain separate.
| Option | Best fit | Price and billing basis | Buying unit |
|---|---|---|---|
| 1. Susan | A managed job instead of another tool | Not published; fixed monthly quote for agreed workflows | Agreed managed job / workflows |
| 2. ChatGPT Business | One assistant across a mixed software stack | USD $20/Standard seat/month billed annually; $25 billed monthly | Named paid seat; minimum 2 paid Standard/Premium seats |
| 3. Microsoft 365 Copilot | Work already in Microsoft 365 | USD $21/user/month list add-on, paid yearly; eligible base plan extra | Licensed user plus eligible Microsoft 365 base plan |
| 4. Gemini in Google Workspace | Work already in Google Workspace | Business Starter USD $7/user/month on annual commitment; $8.40 flexible | Workspace user; full suite subscription, not a separate Gemini add-on |
| 5. Claude for Work (Team) | Document-heavy team analysis | USD $20/Standard member/month billed annually; $25 billed monthly | Named seat, Standard or Premium; Team up to 150 seats |
| 6. Zapier (Zap workflows) | Cross-app Zaps owned by operations | Professional from USD $19.99/account/month billed yearly | Account task allocation; Professional 1 user, Team up to 25 |
| 7. Make | Visual scenarios and branches | Core USD $9/organization/month at 10,000 credits, annual billing | Organization/scenario credits; annual-billing displayed selection |
| 8. n8n | Technical workflows and hosting choice | Starter EUR €20/month billed annually | Production workflow executions, not per-step task billing |
| 9. Lindy | Shared inbox routines and credit controls | USD $29.99/active user/month at 3,000 credits | Every active user adds a paid seat and credits to a shared pool |
| 10. Gumloop | AI orchestration with usage accounting | Pro USD $37/organization/month at 20,000 credits | Shared organization credits; unlimited listed seats/teams |
| 11. HubSpot / Breeze and Agent Hub | AI on existing HubSpot records | Starter Core Seat USD $20/seat/month; feature credits extra | Hub subscription and seats plus feature-specific credits |
| 12. Notion AI (Business) | AI in an existing Notion knowledge base | Business USD $20/member/month billed annually; $24 billed monthly | Workspace member plus optional credit-powered Custom Agents |
| 13. Perplexity Enterprise Pro | Research with inspectable sources | USD $40/seat/month or $400/seat/year | Enterprise seat plus Computer credits |
| 14. Fireflies.ai | Cross-platform meeting capture | Pro USD $10/seat/month billed annually; $18 billed monthly | Workspace paid seat plus limited AI credits / applicable API limits |
| 15. Grammarly | Everyday writing and editing | Pro USD $12/member/month equivalent; $144 annual plan | Member subscription; generative prompt allowance per member/month |
| 16. Salesforce Agentforce | AI on existing Salesforce records | USD $500/100,000 Flex Credits or $2/eligible conversation | Action credits, eligible conversation SKU or licensed employee add-on |
Our first pick
Why Susan is the best choice for most teams
Susan works like a hiring agency for AI employees: bring us the job description, and our team builds the AI employee, runs it and keeps it working, for one price you can plan around.
- One fixed monthly price. For agreed hosted workflows, model usage, hosting, monitoring and fixes are included in one fixed monthly price. Customers using their own model account pay that provider directly. Compare the named providers’ actual pricing and scope in the article’s sourced table.
- Done for you, not do-it-yourself. Bring us the job description and a screen recording of how the work is done today: if a person does it on a computer, Susan can learn it. There is nothing for your team to configure, prompt or maintain, and the first job is running within 48 hours of signing.
- Costs less than another hire. No recruiting, training or turnover, and busy season costs the same. The work keeps moving at night and on weekends, and anything that needs approval waits for your staff.
- Support and fixes are part of the fee. When a portal changes, a login expires or a new exception appears, our team fixes it. That work is included, not billed as extra hours or left to your staff.
- The safest setup: your own cloud and your own keys. Susan can run entirely inside your team's own Microsoft Azure subscription, with your own OpenAI, Anthropic or Azure OpenAI account, and our engineers run it there for you. Your data never has to sit on a vendor's platform: the agents, logs and keys stay in your account, under your rules, and you can see and revoke our access.
- Security from the first day. Each agent works on its own dedicated, hardened computer in a US data center. App passwords stay in a protected layer the agent never sees, people approve anything that reaches a client or changes the books, and every action is logged.
- Keeps the software you already use. Susan works through direct app connections and uses its own screen for portals without one, so there is no new platform to move to.
Susan is our product and a managed alternative to the software seats below. The first-job timeline applies to Susan as we host it; own-cloud deployment depends on your IT team granting access. Requests to an outside model provider follow that provider’s terms.

Key takeaways
- Check the office suite, CRM and knowledge workspace you already pay for before adding another AI subscription.
- A paid seat buys access; it does not establish that an included allowance can handle your workload.
- One request can create several billable actions. Compare each vendor’s meter separately and include retries.
- Name the operator who will maintain permissions, instructions and recovery. Keep consequential actions behind review.
- Hold the purchase when nobody can operate the tool. Consider managed execution only when you need the recurring job run for you.
This guide compares software procurement for an established business. For delegating a continuing role, read AI agents for businesses and AI employees for businesses. For the managed alternative itself, see how Susan works. Do not buy a roster of tools to recreate an employee by accident.
How we chose these 16 options
We compared public product, pricing and help pages using the same questions: what software the business already owns, what the selected plan includes, how usage is billed, and who will operate the configured work. The best-for labels are purchasing judgments, not accuracy scores. The worked budget is a planning example, not a measured vendor benchmark.
- Existing-stack fit: Could the current office suite, CRM or knowledge workspace handle the task without another subscription?
- Purchase terms: What are the currency, seat minimum, billing commitment, included allowance and overage behavior?
- Task boundary: Does the output stop at research or a draft, or does a configured workflow create records and require approvals?
- Operating cost: Who checks output, resolves an expired connection and repairs a partial run?
The selection spans general assistants, office-suite AI, workflow software, CRM features and specialists. These products do different jobs; a meeting recorder and an automation engine are not interchangeable. We excluded a second employee-hiring roster. Use the linked vendor terms to confirm your country, edition and workload-sized tier before purchase.
What adoption figures do, and do not, tell a buyer
Gallup’s July 20, 2026 report found that 47% of employed US adults said their organization had integrated AI to improve productivity, efficiency or quality. The May 6–20 survey covered 22,573 employed adults. That is employee-reported organizational adoption, not the share of businesses buying any particular tool.
In the 2026 American Job Quality Study, 63% of 7,845 US employees who had used AI in their job at least a few times in the past year said it helped them complete tasks faster. Fieldwork ran January 26–March 24. This is a share of respondents reporting faster work, not a measured 63% productivity increase or a forecast for these products.
The US Census Bureau’s Business Trends and Outlook Survey reported a national business AI-use rate of 19.8% for the collection period ending May 3, 2026. Its nationally representative business survey asked about use in any business function over the preceding two weeks. That firm-level measure and Gallup’s employee measures have different populations and questions; they are not interchangeable adoption estimates.
Two figures come from Gallup, so these are not three independent replications. None tells you which subscription your business needs. Measure accepted output and retained operator time in your own limited evaluation.
16 business AI tools and their buying limits
For each option, the four linked facts describe vendor-published capabilities. Fit, limits and the recommended operating owner are our analysis. A vendor’s product support does not automatically maintain your company’s configured workflow.
1. Susan: A managed job instead of another tool
A firm that wants a recurring cross-app job run and maintained for it rather than buying another tool to operate.
Price and buying unit: Not published; fixed monthly quote for agreed workflows. Agreed managed job / workflows.
For agreed hosted workflows, model usage, hosting, monitoring, support and fixes are included. Customers using their own model account pay that provider directly. Agree the job and approvals before comparing quotes. Official product and pricing.
- Completes recurring business work across connected apps and uses its own computer for steps that need a screen. Source.
- Each agent has a dedicated managed computer; the team can watch live, take over and inspect the work afterward. Source.
- Fixed Labs sets up, operates and fixes the agreed workflow when systems change. Source.
- Consequential actions such as sending or posting invoices require staff approval; unusual cases are escalated. Source.
Limit: Needs an agreed job, supported applications, authorized access and a named reviewer. Setup and professional approval responsibilities remain explicit; no provider can promise mistake-free work.
Who operates and repairs it: Fixed Labs owns implementation, monitoring and fixes; the firm owns process policy, authorized access and approvals.
2. ChatGPT Business: One assistant across a mixed software stack
Mixed-stack teams choosing a standard assistant for drafting, research and analysis.
Price and buying unit: USD $20/Standard seat/month billed annually; $25 billed monthly. Named paid seat; minimum 2 paid Standard/Premium seats.
A workspace requires at least two paid Standard/Premium seats. Premium is USD $100/seat/month billed annually or $125 monthly. Usage limits and extra workspace credits apply. API billing is separate; confirm any conditional API-credit benefit instead of assuming unrestricted API use. Official product and pricing. Vendor source 2.
- Self-serve shared organizational workspace with centralized billing and user administration. Source.
- Standard seats include ChatGPT, ChatGPT Work, Codex, Projects, Apps and company knowledge. Source.
- Business data is not used for model training by default. Source.
- SAML SSO is included; SCIM, Compliance API logs and configurable data residency are Enterprise-only in the comparison. Source.
Limit: Users still check outputs and supply context. Connected actions need permissions and operating ownership; Enterprise-only governance is not included merely because a workspace is Business.
Who operates and repairs it: Workspace admin manages users, seats and apps; process owner validates outputs; internal builder maintains custom workflows; OpenAI supports the product.
3. Microsoft 365 Copilot: Work already in Microsoft 365
An organization whose everyday documents, mail and meetings already live in Microsoft 365.
Price and buying unit: USD $21/user/month list add-on, paid yearly; eligible base plan extra. Licensed user plus eligible Microsoft 365 base plan.
This is the Business add-on list price, not a full Microsoft 365 subscription or a temporary offer. Teams features require a Teams license. Advanced usage-based experiences can add charges; verify the selected base edition and billing policy. Official product and pricing.
- Works in Word, Excel, PowerPoint, Outlook and Teams with organizational work context. Source.
- Includes prebuilt research and analysis agents. Source.
- Prompts and responses are not used to train underlying large language models. Source.
- Admins manage usage-based billing and spending policies in Microsoft 365 Cost Management. Source.
Limit: Base licensing and usage charges can exceed the add-on price. Tenant permissions and source-document quality need attention; platform support does not run the business process for you.
Who operates and repairs it: Microsoft 365 tenant admin and internal process/builder owner; Microsoft repairs platform defects.
4. Gemini in Google Workspace: Work already in Google Workspace
Companies working in Gmail, Docs, Sheets and Drive that can use AI already in their suite.
Price and buying unit: Business Starter USD $7/user/month on annual commitment; $8.40 flexible. Workspace user; full suite subscription, not a separate Gemini add-on.
Starter includes Gemini in Gmail and the Gemini app. For broader document and meeting AI, check an edition such as Standard: USD $14/user/month on annual commitment or $16.80 flexible. Annual committed paid seats cannot be reduced until renewal. This is a suite price, not an extra Gemini seat. Official product and pricing. Vendor source 2. Vendor source 3.
- Gemini AI features are included in Workspace Business/Enterprise editions; legacy Gemini Business/Enterprise add-ons are no longer sold. Source.
- Workspace edition determines access to Gmail, Docs, Sheets, Slides, Drive, Chat and Meet AI. Source.
- Existing Workspace data protections apply; content is not used for generative model training outside the domain without permission. Source.
- Admins control Gemini app access and meeting note-taking, but turning off the Gemini app does not switch off every in-app AI feature. Source.
Limit: Use the suite you already operate when it meets the task. Changing office suites to chase an AI headline price adds migration and administration work; controls vary by edition.
Who operates and repairs it: Workspace admin manages service settings and access; document/process owners verify work; Google maintains service.
5. Claude for Work (Team): Document-heavy team analysis
Teams doing writing, analysis and document-heavy work that want a standalone workspace.
Price and buying unit: USD $20/Standard member/month billed annually; $25 billed monthly. Named seat, Standard or Premium; Team up to 150 seats.
Team requires at least two members and supports up to 150 seats. Premium is USD $100/member/month billed annually or $125 monthly. Optional extra usage credits extend limits; one member’s included usage cannot be transferred to another. Official product and pricing. Vendor source 2.
- Team includes centralized access and billing administration. Source.
- Workplace connectors can retrieve context from Google Drive, Gmail, Calendar, GitHub, Microsoft 365 and Slack. Source.
- Team includes projects, collaborative work and enterprise search. Source.
- Usage limits apply individually to each member; reaching one member’s limit does not consume another’s allowance. Source.
Limit: Per-member limits can interrupt a heavy user even when colleagues have unused capacity. Confirm governance in the actual contract; Team should not be assumed to include Enterprise audit and provisioning controls.
Who operates and repairs it: Team owner controls access/spend and connectors; the firm checks output and maintains custom skills; Anthropic maintains product.
6. Zapier (Zap workflows): Cross-app Zaps owned by operations
A nondeveloper operations owner maintaining a workflow across a mixed SaaS stack.
Price and buying unit: Professional from USD $19.99/account/month billed yearly. Account task allocation; Professional 1 user, Team up to 25.
The entry Professional tier has 750 monthly tasks and one user. Team starts at USD $69/account/month billed yearly with up to 25 users. Select the task tier for the real workload; Zapier Agents activities are a separate meter. Official product and pricing.
- A Zap combines one trigger with one or more configured actions. Source.
- Successful actions consume tasks; triggers/polling are not charged as tasks in the FAQ. Source.
- AI/model tier, code runtime and connector type can change task rates; all relevant products draw on the account pool. Source.
- Team supports shared Zaps, folders, shared connections and SAML SSO. Source.
Limit: Additional billable steps multiply consumption. No-code still requires someone to repair mappings and expired connections. This row concerns Zap workflows, not hiring a Zapier Agent.
Who operates and repairs it: The firm’s automation owner or contracted implementer repairs mappings, credentials and workflow logic; Zapier maintains the service/connectors.
7. Make: Visual scenarios and branches
Teams wanting explicit visual branches, transformations and error handling.
Price and buying unit: Core USD $9/organization/month at 10,000 credits, annual billing. Organization/scenario credits; annual-billing displayed selection.
The displayed Core price uses annual billing and 10,000 monthly credits. Pro is USD $16/month and Teams USD $29/month at that selection. Most module actions use one credit; AI and advanced modules can cost more. Custom model connections also have the provider’s token bill. Official product and pricing. Vendor source 2. Vendor source 3.
- A scenario defines how apps exchange and transform data. Source.
- Module actions consume credits, so one scenario run can consume several or many credits. Source.
- Core includes unlimited active scenarios, minute-level scheduling and Make API access. Source.
- When credits run out, scenarios pause until credits are added; queued webhooks are subject to queue capacity. Source.
Limit: An inexpensive organization plan is not a price per completed request. Iterators, polling and dynamic AI consumption can change the bill; keep an owner for incomplete runs.
Who operates and repairs it: Internal automation builder/partner owns scenario logic, errors, reconnects and rollback; Make owns platform availability.
8. n8n: Technical workflows and hosting choice
A firm with technical ownership that wants custom logic and a choice of hosted versus self-hosted operation.
Price and buying unit: Starter EUR €20/month billed annually. Production workflow executions, not per-step task billing.
Pro is EUR €50/month billed annually. Starter has one shared project and five concurrent executions. Confirm the selected production-execution allowance before purchase. Assistant credits are a different allowance. Self-hosted Community Edition excludes hosting, maintenance and model-provider costs. Official product and pricing.
- Pricing is based on monthly workflow executions; plans list unlimited users/workflows and integrations. Source.
- An execution is one run of the full workflow rather than an individual action step. Source.
- Includes AI nodes, JS/Python steps, custom API requests and human approval for tool calls. Source.
- Offers n8n-hosted cloud and self-hosted options; Community Edition is available on GitHub. Source.
Limit: Self-hosting adds updates, backups, security and recovery to the operator’s responsibilities. A published starting price does not establish that its quota covers the sample workload below.
Who operates and repairs it: Internal technical owner/implementation partner owns workflow behavior and any self-hosted runtime; n8n owns its cloud platform.
9. Lindy: Shared inbox routines and credit controls
An operations owner buying a configurable assistant around mail, meetings and team requests.
Price and buying unit: USD $29.99/active user/month at 3,000 credits. Every active user adds a paid seat and credits to a shared pool.
The selector also lists 5,000 credits at USD $49.99/active user/month; public entry views can differ, so confirm checkout. Slack mentions can create paid seats after the trial. Monthly credits expire; top-ups cost USD $10/1,000 credits and remain until used. Confirm computer-use entitlement for the selected tier. Official product and pricing.
- Supports scheduled routines, inbox work, meeting notes and shared workspace context. Source.
- Credits are pooled across the team and admins can allocate credits per seat. Source.
- Outside-impact actions such as sending email, posting or updating tickets wait for approval according to its pricing FAQ. Source.
- If the pool is exhausted, credit-using actions pause until reset or authorized top-up. Source.
Limit: Credit use varies with task and model, while active-user billing can expand the seat count. Buy a defined routine and allowance, not a promise of unlimited inbox labor.
Who operates and repairs it: Workspace admin and the firm’s assigned routine owner handle permissions, instructions, review and recovery; Lindy supports service.
10. Gumloop: AI orchestration with usage accounting
An AI operations owner who needs model/tool orchestration and detailed usage accounting.
Price and buying unit: Pro USD $37/organization/month at 20,000 credits. Shared organization credits; unlimited listed seats/teams.
Credits cover model, connector and compute consumption plus the documented 8% orchestration fee. Overage is USD $0.005/credit. With your own model key you pay that provider directly and the documented orchestration fee changes to 16%; this is a different cost configuration. Official product and pricing. Vendor source 2.
- Pro lists 20,000 monthly credits and unlimited agents and seats. Source.
- Successful connector calls cost at least 1 credit plus any extra tool charge. Source.
- Compute is billed at 5 credits per active session-minute, with rounding/minimum rules. Source.
- Pro credits do not roll over; overage can be enabled with a configurable cap and agents stop at the cap. Source.
Limit: A prompt can incur model, tool-call and compute charges. The current product spans agents and orchestration; older builder-only descriptions and historical plan prices are not a sound purchasing basis.
Who operates and repairs it: The firm’s agent/workflow owner maintains instructions, connectors, permissions and acceptance checks; vendor fixes the platform. Enterprise embedded help must be scoped.
11. HubSpot / Breeze and Agent Hub: AI on existing HubSpot records
A firm already operating sales/service/marketing inside HubSpot that wants AI on those customer records.
Price and buying unit: Starter Core Seat USD $20/seat/month; feature credits extra. Hub subscription and seats plus feature-specific credits.
This catalog Core Seat includes Breeze Assistant within purchased products, not every HubSpot hub for USD $20. Additional credits cost USD $0.010 each or $10/1,000. The catalog lists 50 credits per customer-agent text resolution and 10 per data-agent record/prompt or AI workflow action. Flexible seats-and-credits contracts differ. Official product and pricing. Vendor source 2.
- Breeze Assistant is part of the Core Seat’s access to purchased HubSpot products. Source.
- Agent Hub builds and manages marketing, sales and service agents using live CRM context. Source.
- Included credits reset monthly; unused credits do not roll over and allowances are not additive across products. Source.
- The catalog distinguishes per-action credit charges from recurring monthly credit charges. Source.
Limit: Seat type, purchased hubs and credit rates determine what is included. Configure permissions and CRM data first; an included AI assistant does not grant every paid feature.
Who operates and repairs it: HubSpot super admin/RevOps owner maintains records, permissions, knowledge and workflows; HubSpot supports its platform.
12. Notion AI (Business): AI in an existing Notion knowledge base
A team that already keeps procedures, projects and shared knowledge in Notion.
Price and buying unit: Business USD $20/member/month billed annually; $24 billed monthly. Workspace member plus optional credit-powered Custom Agents.
Annual Business totals USD $240/member/year. Custom Agents use separate Notion credits at USD $10/1,000 monthly credits. Free/Plus offer an AI trial rather than the full continuing Business entitlement. Enterprise governance is a separate purchase. Official product and pricing. Vendor source 2.
- Business includes Notion Agent, AI Meeting Notes and Enterprise Search (beta). Source.
- Business includes SAML SSO, private teamspaces and granular database permissions. Source.
- Enterprise adds SCIM, audit logs and zero data retention with LLM providers. Source.
- Custom Agents use separately purchased Notion credits rather than unlimited seat-bundled use. Source.
Limit: Zero retention with model providers is an Enterprise feature, not a blanket Business promise. A generated note still needs checking and a deliberate handoff into other record systems.
Who operates and repairs it: Workspace admin manages permissions/connections; wiki/process owner keeps source content current and verifies updates; Notion maintains product.
13. Perplexity Enterprise Pro: Research with inspectable sources
Analysts who repeatedly need research with inspectable sources.
Price and buying unit: USD $40/seat/month or $400/seat/year. Enterprise seat plus Computer credits.
The annual charge is USD $400 per seat, not USD $40 for twelve months. Each seat includes 500 monthly Computer credits in the captured plan; API use is separate. Removing a user frees a seat, but reducing purchased seats takes effect at the next billing period. Official product and pricing. Vendor source 2. Vendor source 3.
- Enterprise Pro combines web, file and app search with cited answers. Source.
- Includes 500 monthly Computer credits per seat on the captured plan. Source.
- Enterprise page states customer data is not used to train models. Source.
- The enterprise offering lists admin permissions and SSO; additional SCIM, audit-log and retention controls depend on the contract. Source.
Limit: Citations are a route to checking a claim, not a guarantee it is correct. Confirm contract-specific SCIM, audit and retention controls before relying on them.
Who operates and repairs it: Enterprise admin manages source access; analyst owns fact checking and research acceptance; vendor maintains search/product.
14. Fireflies.ai: Cross-platform meeting capture
A business needing a separate cross-platform meeting record rather than another general assistant.
Price and buying unit: Pro USD $10/seat/month billed annually; $18 billed monthly. Workspace paid seat plus limited AI credits / applicable API limits.
Business is USD $19/seat/month billed annually or $29 monthly. Pro has 8,000 storage minutes per seat and 20 AI credits; Business has unlimited storage and 30 AI credits. Unlimited transcription/summaries do not mean unlimited AI-credit actions or API volume. Enterprise is USD $39/seat/month, annual only. Official product and pricing.
- Records/transcribes supported meetings across Zoom, Google Meet, Teams and other platforms. Source.
- Pro includes unlimited AI summaries, downloads and integrations. Source.
- Business adds conversation intelligence and team analytics. Source.
- Enterprise adds SSO/SCIM, custom data retention and API audit logs. Source.
Limit: Check the meeting features in your current office suite first. Consent, sharing and retention remain your responsibility, and a meeting summary is not an approved follow-up record.
Who operates and repairs it: Meeting/workspace admin owns invitations, sharing, retention and CRM mappings; meeting owner checks decisions and follow-ups; Fireflies maintains service.
15. Grammarly: Everyday writing and editing
A team primarily trying to improve everyday written communication rather than execute a multi-app process.
Price and buying unit: Pro USD $12/member/month equivalent; $144 annual plan. Member subscription; generative prompt allowance per member/month.
The USD $12 figure is an annual equivalent, not a verified monthly checkout offer. Pro includes 2,000 generative prompts/member/month; Free has 100. Confirm team eligibility and checkout terms; Enterprise pricing is unpublished. Official product and pricing.
- Pro offers full-sentence rewriting and tone adjustment. Source.
- Pro includes 2,000 generative AI prompts per member per month. Source.
- Pro adds team functionality previously in the Business plan. Source.
- Enterprise adds custom roles/permissions and data-loss prevention. Source.
Limit: Writing edits do not verify facts or complete cross-app work. Review sensitive-text controls and which applications receive assistance before a company-wide rollout.
Who operates and repairs it: Firm’s IT/workspace admin manages rollout and rules; document authors approve edits; Grammarly maintains product.
16. Salesforce Agentforce: AI on existing Salesforce records
A business already committed to Salesforce records and processes.
Price and buying unit: USD $500/100,000 Flex Credits or $2/eligible conversation. Action credits, eligible conversation SKU or licensed employee add-on.
A standard action consumes 20 Flex Credits; Voice consumes 30. The USD $5/user/month access license still requires usage credits. Employee add-ons include a USD $125/user/month option. Base Salesforce licenses and other services can apply; Flex Credits and Conversations cannot be combined in one org. Official product and pricing.
- Agentforce supports both customer-facing and employee-facing agents. Source.
- Agentforce Builder and Prompt Builder are part of the platform offering. Source.
- Digital Wallet provides granular consumption monitoring. Source.
- Unused Flex Credits do not roll over into subsequent subscription terms. Source.
Limit: Choose it around an existing Salesforce process, not as a reason to buy a CRM. Action design, record access and consumption need a Salesforce administrator.
Who operates and repairs it: Salesforce admin/architect and business process owner maintain actions, permissions and recovery; Salesforce maintains platform.
A worked budget: 200 weekly requests are not 200 credits
Take an internal request that arrives through a form. Staff want it classified, matched to an authorized source, added to an internal task list and turned into an unsent draft for review. A general assistant may help with the content; an automation platform needs the steps configured. Neither access to a model nor a meeting transcript proves that the full request has been completed.
Assume 200 requests per week for four budgeting weeks: 200 × 4 = 800 requests. This is not an average calendar month. Start with the following assumed design, then replace the assumptions with actual billing events from representative runs.
| Software meter | Assumed design | Four-week arithmetic | Purchase constraint |
|---|---|---|---|
| ChatGPT or Claude seats | Ten staff each need a paid seat | 10 named seats | Request length and model limits still need checking |
| Zapier tasks | Three successful billable actions per request | 800 × 3 = 2,400 base actions | Entry 750-task tier is too small; allow for multipliers and retries |
| Make credits | Five one-credit module actions per request | 800 × 5 = 4,000 base credits | AI tokens, iterators and retries can add charges |
| n8n executions | One production run per request | 800 × 1 = 800 executions | Confirm selected allowance and extra executions from retries/sub-workflows |
| Lindy or Gumloop | Model, tools and compute vary | Measure representative runs before setting a cap | No fixed credit conversion is assumed |
| HubSpot or Agentforce | The specific action or conversation SKU | Map each event to that vendor’s rate sheet | A conversation, action and resolution are different units |
| Susan | An agreed managed job | Request a fixed monthly scope and quote | No fee inferred from seats or this sample |
Those calculations describe allowance demand, not equal product performance. Make’s 10,000-credit selection exceeds the assumed 4,000 base events, but that alone does not establish headroom for variable AI consumption. We have not established n8n Starter’s selected production quota. Price the workload-sized tier, base apps and model costs before choosing.
Add the human work that remains
For this planning example, assume 85% ordinary requests need two minutes of review, 15% exceptions need seven minutes, and your operator spends eight hours on permissions, monitoring and recovery during the four-week period. These are budget inputs, not observed acceptance rates or measured product results.
- Ordinary requests: 800 × 85% = 680; 680 × 2 ÷ 60 = 22.67 review hours.
- Exceptions: 800 × 15% = 120; 120 × 7 ÷ 60 = 14 review hours.
- Total retained time: 22.6667 + 14 + 8 = 44.6667 hours, rounded only for display.
- At an assumed loaded internal cost of USD $40/hour: 44.6667 × $40 = USD $1,786.67.
- Ten ChatGPT Business Standard seats on monthly billing: 10 × USD $25 = USD $250.
- Subscription plus assumed retained human budget: USD $250 + $1,786.67 = USD $2,036.67, before extra usage, initial setup, base applications and taxes.
This assistant-budget line does not price an automated completed job. Do not carry the same review assumptions across every tool, subtract them from a made-up salary or call the result ROI. Record accepted requests, actual review minutes and recovery work during your evaluation, then recalculate.
Need the job run rather than a tool to administer? Book a walkthrough with the current process and its expected volume.
How to choose AI software without buying duplicate seats
- Inventory the current stack. List the office suite, CRM, note-taking and knowledge subscriptions you already own. Ask which edition includes the missing capability and what an upgrade adds.
- Define the output. Write down whether the tool must produce a draft, create an internal task, update a record or send a message. Keep the final two behind appropriate approval rules.
- Quote the buying unit. Count paid users, successful billable actions, module credits or executions. Confirm whether allowances pool, expire, pause work or create overages.
- Name the operator. Assign a person who can reconnect apps, edit rules and recover a partial run. Assign a reviewer and access administrator too; vendor support is a separate responsibility.
- Run a bounded evaluation. Require the same ordinary request, missing source, expired permission, exhausted allowance and duplicate input from each serious candidate. Keep the source, proposed output and reviewer decision. These are acceptance requirements to demonstrate, not checks these vendors have already passed.
- Make a buy or hold decision. Continue only if the accepted output and retained work justify the whole operating bill. Hold if the current suite already covers the need, no operator is available or the vendor cannot explain the selected allowance.
Business data, mistakes and other buying objections
Is our business data safe?
Safety depends on the selected plan, source permissions and deployment, not the word “business” on a price card. Check model-training terms separately from retention, who can see shared records, and how administrators revoke access. Keep customer information out of an evaluation until your business approves the accounts and data handling.
Susan can run inside your own cloud with your own model keys, operated by our engineers through access your IT team grants and can revoke. Hosted agents each have a dedicated managed computer in a US data center, and credentials stay in a protected layer. An outside model provider still processes requests under its terms; no deployment choice removes your responsibility to approve data use.
Will it make mistakes?
Plan for incorrect output and partial completion. Require a source reference, a checkable destination and an exception owner. Keep consequential actions, including external sending and financial changes, approval-gated. For Susan, staff approve those steps, work is checked and read back, and unusual cases are escalated. A successful demonstration is not permission to remove review.
Why pay for a specialist if our suite includes AI?
Pay only for a gap your existing suite fails to close. Cross-platform meeting capture, research source retrieval or a configured multi-app workflow can justify another tool. A second assistant producing a similar draft usually needs a stronger case. Compare incremental cost and operator effort; a low entry price is not a reason to add a new system.
Does self-hosting remove the running cost?
No. Self-hosting can change control and deployment, but your team still pays for compute, updates, backups, monitoring and any external models. In n8n, self-hosted software and hosted production quotas are separate procurement choices. With Susan in your own cloud and model account, your business retains those accounts and their bills while our team operates the agreed job.
Frequently asked questions
What are the best AI tools for businesses?
Start with the software your business already uses. Copilot fits Microsoft 365 work; Gemini fits Google Workspace. ChatGPT Business and Claude Team are general assistant choices, while Zapier, Make and n8n connect configured workflows. Add a specialist only for an unmet need. Susan is our managed alternative when you want a defined recurring job implemented, operated and maintained for you instead of another staff-operated subscription.
How much should a small business budget for AI tools?
Budget for the selected seats and actual usage allowance, then add setup, review and maintenance. In this comparison, ChatGPT Business Standard is USD $25 per seat on monthly billing, with at least two paid seats. Lower annual equivalents require a longer commitment. Automation tools can charge for several actions inside one request, so an advertised entry price is not a quote for your full workload.
Should we choose Copilot, Gemini or a separate AI assistant?
Choose the assistant in your existing office suite first if it handles the work and its permissions meet your requirements. Copilot’s add-on needs an eligible Microsoft 365 plan; Gemini features depend on the Workspace edition. Consider ChatGPT Business or Claude Team for a specific gap in cross-stack analysis or document work. Compare the same sample task before buying overlapping seats for the whole company.
Are AI credits the same across business tools?
No. Vendors meter different events. Zapier counts successful billable actions, Make charges module credits with variable AI rates, and n8n prices production workflow executions. Lindy and Gumloop use their own model, tool and compute rules. A request can therefore consume several units, and retries can add more. Convert your process into each vendor’s billable events separately; never compare two credit counts as equal work.
Can we safely connect AI software to business records?
Use a business-approved account, limit source access and agree which actions require review before connecting records. Check training terms, retention, sharing and the ability to revoke permissions in the selected plan. A no-training promise does not mean zero retention. Keep sending, financial changes and other consequential actions approval-gated, and require a way to inspect what changed and recover from a partial or duplicate write.
When should we choose a managed service instead of AI software?
Choose managed operation when the requirement is a recurring job and your team cannot own configuration, monitoring and recovery. Susan is our service for that decision, with one fixed monthly quote for agreed hosted workflows, including model usage, hosting, monitoring, support and fixes. Customers using their own model account pay that provider directly. The agreed job needs a supported source, authorized destinations and a named reviewer. Confirm these before choosing a fixed scope; professional decisions remain with the firm.
Bring one workflow before buying another subscription
Bring the software you use, the output you need and who handles it today. We will discuss whether staff-operated software is enough or whether a managed job is the better fit.
Book a walkthrough