What should you ask a graph consulting partner before hiring them?

Ask them to walk through how they'd model permissions before they mention the database vendor. A partner who leads with 'we'll pick the right platform for your constraints' rather than a fixed vendor answer is usually further along than one who opens with a product pitch.

What is the single question that separates architects from implementers?

Ask how they'd design permission enforcement for a graph with genuinely sensitive, unevenly accessible content. An implementer describes filtering results after retrieval. An architect describes scoping every traversal from the requesting user outward, so inaccessible data is never reached in the first place: because they've already seen what post filtering costs you in production.

Should pricing or team seniority weigh more heavily in the decision?

Weigh seniority per seat over headcount. A small team of engineers who have each shipped production graph systems for years will out deliver a larger team where the people you meet in the sales process aren't the people who do the build. Ask specifically who is staffed on the engagement day to day, by name, before signing.

Enterprise Knowledge Graphs

How to Choose a Neo4j Enterprise Partner

Q: What references should you actually check?

Ask to speak with the technical lead of a comparable engagement, not just a sponsor or an account manager: and ask that person specifically what broke during the build and how it was handled. A partner with nothing that ever broke is either withholding or hasn't shipped enough production systems to have a real answer.

Q: How should a production pilot be structured?

Fixed scope, a documented and measurable success target agreed before work starts, and a short enough timeline that both sides find out quickly whether the fit is real. A pilot with a vague success definition, or one that can quietly expand in scope mid engagement, isn't really a pilot: it's the first phase of an open ended contract.

Published 2026-08-31 · Agentic Giants

TL;DR

Agentic Giants' engineers lay out the questions worth asking before hiring a graph consulting partner: how they'd design permissions before they mention a vendor, what a reference's technical lead says actually broke during the build, and how a fixed scope pilot with a documented success target should be structured before a full engagement begins. Team seniority per seat matters more than headcount.

What should you ask a graph consulting partner?

Lead with permissions, not with platform. Ask a candidate partner to walk through how they'd design access control for a graph containing content with genuinely uneven sensitivity: before the conversation ever gets to which database vendor or which framework they favor. A partner whose first instinct is “we'll choose the right platform for your constraints” rather than opening with a fixed product pitch is usually further along the maturity curve than one who leads with a tool.

What is the one question that separates architects from implementers?

Ask specifically how they'd enforce permissions in a system where different users genuinely can't see the same data. An implementer will describe retrieving broadly and filtering results afterward: which sounds reasonable and is quietly unsafe, because it silently degrades answer quality for anyone with narrower access and can leak the existence of hidden content through counts and summaries even when nothing restricted is ever directly shown. An architect will describe scoping every retrieval to traverse outward from the requesting user, so that inaccessible data is never reached in the first place: because they've already watched the alternative fail somewhere.

What references should you actually check?

Ask to speak directly with the technical lead on a comparable prior engagement: not only a sponsor or an account manager, who will naturally speak to the relationship rather than the engineering. And ask that person a specific, pointed question: what broke during the build, and how was it handled? A partner with a real, detailed answer to that question has actually shipped production systems. A partner whose references have nothing that ever went wrong is either not being candid, or hasn't shipped enough to have encountered a real failure yet, and either one is worth knowing before you sign.

How should a production pilot be structured?

Three things, non negotiably: fixed scope, a success target that is specific and measurable and agreed in writing before work starts, and a timeline short enough that both sides learn quickly whether the fit is real. A pilot without a clearly documented success definition, or one whose scope can quietly expand mid engagement without a new agreement, isn't actually a pilot: it's the opening phase of an open ended contract wearing a pilot's name. If a prospective partner resists putting the target in writing before the engagement starts, that reluctance is itself useful information.

Does team seniority matter more than headcount?

In graph and GraphRAG work specifically, yes. A small team where every engineer has actually shipped production graph systems will consistently out deliver a larger team assembled around the engagement, where the senior people you meet during the sales process aren't the people who show up to do the build. Ask, by name, who is staffed on the engagement day to day (not who's available for the kickoff call) before you sign anything.

Keep reading

Pillar

How to Choose a Neo4j Enterprise Partner

What should you ask a graph consulting partner?

What is the one question that separates architects from implementers?

What references should you actually check?

How should a production pilot be structured?

Does team seniority matter more than headcount?

Keep reading

Enterprise GraphRAG: implementation patterns →

Neo4j Enterprise: production hardening →

Our team →

RYVYL: NASDAQ fintech case study →