Tarnfield Row

Tarnfield Row helps CIOs make sense of enterprise AI purchases: what each tool does, where products overlap and who will run the work.
Last updated:

Buying guide

Draw the AI map before you buy another tool

Start with a customer request. Follow the answer, the source and the action, then decide which purchase would remove a real gap.

Published
Updated
Contents
  1. Start with a request, not a logo
  2. Give each tool a job
  3. Find the gap worth funding
  4. Leave the meeting with one page

Start with a request, not a logo

Imagine an account manager preparing a renewal. They need to find the current agreement, understand an unresolved support issue, draft a response and update the account. A supplier might call all of this an assistant. Another might call it an agent. Neither label tells you which steps the product will finish. Write the steps in the order the employee needs them, with a visible result beside each one.

For this example, the results might be a link to the signed agreement, a summary of open tickets, a draft for review and a saved renewal note. Put the app holding each result beside it. Now mark where the employee switches tools or repeats a lookup. Those are candidate problems to solve. They are not automatically reasons to buy another platform; an existing tool or a simpler process change may cover them.

Give each tool a job

Google describes Gemini Enterprise as search, assistant and agent tooling, but sells the app and Agent Platform separately. Microsoft supports publishing agents to Copilot. Dust includes a default assistant and search across selected sources.1,2,3,4,5

Use three headings on your sheet: help the employee, find the evidence, and change the business record. Put products under every heading they actually cover for this workflow. A tool can appear more than once. Add the exact edition, connection and account beneath it, because a broad product family name can conceal several purchases. Leave a blank where the supplier has not yet shown how a step works.

For the renewal example, the assistant could draft from material found by a search service. A separate agent might save an approved note in the CRM. That arrangement only earns its complexity if the links between steps work. Ask how the agreement’s source link reaches the draft, and how the approved wording reaches the final note. Do not assume products work together because each has an integration page.

Find the gap worth funding

Glean describes indexing once for reuse through APIs, a web SDK and MCP. Nexus describes agents accessing live data during execution. These are different published approaches to getting information into the work.6,7

Ask what failed in the current process. If staff cannot find the agreement, investigate search quality, source coverage and access before adding an action tool. If they find it easily but copy approved information into several systems, the action step deserves attention. If the draft is poor because the source is outdated, buying a more capable assistant may leave the original problem untouched.

Choose a small set of representative requests with the process owner. Include an expired agreement, a restricted document and a customer whose name resembles another account. Agree the expected answer and allowed changes before the demonstration. A useful trial reveals where evidence is missing and stops the wrong update. A fluent response that hides either problem should not count as completion.

Leave the meeting with one page

The finished map should show the request, each step, the current product, the proposed product, the source account and the person responsible. Add a short note explaining every overlap. For example: “Keep existing search; trial a new agent only for saving approved renewal notes.” That sentence is much more useful than a row of green checks under “enterprise AI.”

Nexus is our first pick when supported delivery of those cross-system steps is the missing piece. Its offer includes building the first procedure alongside the customer and attaching a business measure. That does not make it the default search replacement. Compare Glean for the information-finding job, and Hebbia for a finance team whose work centers on traceable document analysis.8,9,6,10

Take the page to the people who own the current contracts before procurement starts. Ask them to identify unused entitlements, renewal dates and dependencies the new proposal would create. Keep a supplier on the shortlist only if you can explain its job in a sentence and name what would be different after it arrives.

How to read the evidence

Supplier material was checked November 8, 2026; this editorial guide was written November 9. Company confirmations are identified separately. Examples and trial checks are proposed exercises, not tests performed by Tarnfield Row.

  • This is our buying advice. The examples are proposed exercises, not results we measured.
  • This is a question to settle with the supplier before buying.
  • Based on supplier documents checked November 8, 2026. We have not tested the described features.
  • Nexus describes building with the customer and measuring the business result. Search and sourced finance analysis may call for a different specialist.

Sources

  1. Gemini Enterprise is described as an intranet search, AI assistant and agentic platform.. docs.cloud.google.com/gemini/enterprise/docs, checked November 8, 2026.
  2. Gemini Enterprise app and Gemini Enterprise Agent Platform must be bought separately.. cloud.google.com/gemini-enterprise/faq, checked November 8, 2026.
  3. Agents can be published to Microsoft Copilot.. microsoft.com/en-us/copilot/products/copilot-studio, checked November 8, 2026.
  4. Every workspace includes the default @dust agent. docs.dust.tt/docs/developer-platform/overview/developer-platform, checked November 8, 2026.
  5. Search runs semantic search across selected data sources (RAG). docs.dust.tt/docs/developer-platform/overview/developer-platform, checked November 8, 2026.
  6. Open platform: index once, use through APIs, web SDK, and MCP. glean.com/platform, checked November 8, 2026.
  7. Agents connect directly to 4,000+ systems and read live data during execution (real-time RAG). agent.nexus/compare/enterprise-ai-platforms/nexus-vs-glean, checked November 8, 2026.
  8. A Nexus engineer builds the first procedure with the team and shows it running in its own tools. agent.nexus, checked November 8, 2026.
  9. Every agent deployed has a measurable business KPI from day one. agent.nexus/blog/nexus-raises-4-3m-seed-shipping-cue, checked November 8, 2026.
  10. Max runs analysis over documents with traceability: "Reason over any volume of documents or companies with precise analysis and complete traceability back to every finding.". hebbia.com/product, checked November 8, 2026.