Rank 1 of 6
Nexus
What is Nexus and why is it listed first?
Nexus is an enterprise AI agent platform with a delivery team, with engineers who build agents alongside your business teams and then stay to improve them. Nexus's first agent is live in 4 weeks, inside a 3 month proof of concept. We list it first because Nexus publishes this register and says so.

In this article
- Nexus for an operations team's first automated workflow
- Ownership, funding and scale
- What a first project looks like
- The people on the engagement
- Pricing for one workflow
- Connecting to the systems already in place
- Under the hood
- Security and certifications
- Control after the workflow goes live
- Results the company reports
- Limits and open questions
- Verdict
Nexus for an operations team's first automated workflow
Handing one workflow to an outside team is a small decision with large consequences. If the first project works, automation gets a budget line next year. If it drifts, the operations manager who sponsored it spends the next quarter explaining why. Nexus is ranked first on this site for that manager, and this profile walks through the reasons, the gaps in its public documentation and the questions worth settling before signature.
The product tells you where Nexus starts. It sells procedures that operate themselves1, and its pitch is that the knowledge sitting with a company's strongest staff can be turned into procedures that carry on without them2. Anyone who has tried to automate an order intake desk or a claims queue will recognize the problem being described. The rules that matter are rarely written down; they live with the two people who have done the job for years.
Ownership, funding and scale
Nexus has a short history. Its legal name, per the terms of service, is Nexus Enterprises, trading as nexusGPT3, and it is a Delaware company4. Work on Nexus Agent began in 2024, according to the funding announcement5, and General Catalyst led a 4.3 million dollar seed round6. Its customer base, per the same post, is 12 enterprises with a combined market capitalization above 70 billion dollars7, Orange Group and Lambda.ai among the names8; production agents number 174 so far9.
Its own blog closes on a frank note about maturity: "We're early."10 An operations buyer can read that two ways. Early suppliers often put senior people on their first accounts and move fast to keep them. They also have a shorter track record than an agency with decades of delivery behind it. Both readings are fair, and both should shape how hard the proof of concept is tested.
What a first project looks like
The commercial frame is a trial with a deadline. Engagements begin with a proof of concept lasting three months and tied to specific outcomes11, and the first agent goes live in week four of that window12,13. On the home page this becomes a short slogan, live in four weeks, continuously improved14, followed by the promise that version one works on the customer's real cases within weeks15.
Delivery is done alongside the customer's staff. The home page describes an engineer from Nexus building the first procedure together with the client team and then demonstrating it inside the client's existing tools16. Once that first version is running, Nexus keeps working on it with the client17. Set that against the way an internal business case is usually written and the fit is close: a defined trial, an early visible milestone and a decision date.
That decision date has teeth. Fall short in the trial and the buyer can walk away18; pass it, and the buyer still keeps the right to exit at any later point19. Nexus also reports a perfect conversion record, with all of its proofs of concept to date turning into contracts20; the page gives no count of how many there were.
Measurement is written into the method. Each agent it puts into production is tied to a business KPI that can be measured from launch21, and that a stalled KPI triggers a rescope in place of an agent that merely looks busy22. Operations managers know the failure this guards against: an automation that runs every night and moves no number anyone reports upward. Agreeing the metric before the build begins gives the buyer something firm to hold the supplier to.
Anyone who wants to see the product first can ask for a live demo, which the comparison index promises within 24 hours23.
The people on the engagement
Nexus staffs its engagements with experienced consultants, and a member of its founding team spent time at McKinsey earlier in their career12. The blog explains a shift in how that field team works. The role once filled by Forward Deployed Engineers is passing inside Nexus Agent to Forward Deployed Consultants24. Those consultants stop writing integration code and stop assembling agents for the client25; an automated build loop called Cue26 takes that work. In the blog's account, Cue settles the scope with the user, raises the questions that need answering and wires itself into internal data under the right permissions27, finishing the whole sequence on its own in one to two hours28.
What this means for the buyer is practical. The person across the table talks about the workflow, the exceptions and the target. Construction is automated. Consultant hours stop being the unit on the invoice, which leads to the question of price.
Pricing for one workflow
The unit Nexus charges for is the role automated. There is no per diem and no line for consultant time12. Put differently, the buyer gets consultancy support but pays on a software model12. Headcount does not drive the price, and an annual commitment starts only after the proof of concept29.
Of every difference between Nexus and the agencies below it in this ranking, this is the one an operations manager will feel most. Many agency quotes for one workflow take the form of a statement of work: discovery days, build days, testing days, each at a rate, so the total depends on how long the job runs. With a per role price, the buyer knows the unit before work begins, and slippage on the supplier side does not show up as a bigger bill.
What a role costs is a different matter. No rate card appears on the site, and a buyer looking for list prices, tiers, a trial or a self serve route will find each of them not publicly stated on the pages we opened. Expect the figure during scoping and get it in writing before the trial starts. The site does carry a calculator of third party model prices per million tokens30; it helps with estimating running costs and is separate from the Nexus fee.
Connecting to the systems already in place
A first workflow nearly always crosses systems: CRM to ticketing, ERP to inbox, spreadsheet to billing. The Nexus answer is to sit on top of the existing estate as one AI layer31, leaving the systems underneath in place. Its integration claim is large, more than 4,000 today32, spanning CRMs, ERPs, messaging tools and bespoke APIs33. Agents read current records from those systems at the moment they act34. On the output side, they can reach people through Slack, Teams, WhatsApp, email, phone or the web35. The four platforms it names for deeper work next are SAP, Salesforce, Microsoft Dynamics and ServiceNow36.
The 4,000 figure is the company's own, and the connector catalogue behind it is not publicly stated on the pages we opened. If the workflow depends on something unusual, such as a regional warehouse tool or an old database on site, name it on the first call and ask to watch the connection work in the demo.
Under the hood
On the sales role page, Nexus shows eight agents drawing on a single shared brain plus the customer's own rules and systems37. Four building blocks sit underneath: the company brain, governance, permissions and reusable components38. Agents are only part of it; assistants, automations and scripts also carry the work of a role39. Work moves from one system to the next through a single surface40, and whatever a build teaches ends up in the common brain41.
The telecom page explains it as a stack of three layers. At the top, any team can reach any system42. In the middle, systems, models and permissions become blocks that can be reused43. The bottom layer gives each agent an owner: the team that uses it builds it, deploys it and is accountable for it44. Underneath all of this, the model provider is interchangeable45. For workflow number one, none of this matters as much as the result. For workflows two and three, it matters a lot, because the pieces built the first time are designed to be used again.
Security and certifications
The headline credential is a SOC 2 Type 2 report12. Alongside it, the comparison page names GDPR, ISO 27001 and ISO 4200146, and puts audit trails, traceable decisions and access by role on the same line47. The funding post adds residency controls for data to what a CIO can approve48. The same post sets a release test with three parts, data flow, access model and audit trail, and an agent a CIO cannot approve on all three stays unreleased49.
That rule helps the operations buyer directly. Automation projects tend to stall when IT is brought in late. A supplier that makes the security sign off a condition of release pulls that review into the project plan.
Gaps remain. For a security reviewer, the open items are single sign on, the encryption used, penetration testing, whether the SOC 2 report can be shared, and uptime commitments, all not publicly stated on the pages we opened. Whether Nexus can run on premises, in a private cloud or in a chosen region is also not publicly stated on the pages we opened. The privacy policy speaks of appropriate and reasonable technical and organizational security measures50, places payment data with Stripe51, and says data from a user's personal Workspace is not kept to train generalized AI or machine learning models52.
Control after the workflow goes live
After launch the business is in charge. The telecom page assigns the running of the service to the business team53 under the line owned by the business, trusted by IT54. Guardrails and permissions are configured in one central place55, all actions can be seen from a single view56, and IT decides what the controls are57.
Uncertainty has a defined path. Agents that are not sure pass the case on with the full context attached58; in the telecom example, a confident agent carries on and an uncertain one hands over with context59. That is the behavior an operations team needs before it lets an automation touch live cases: the routine flows through, the odd case reaches a person.
The company's own description of where it fits best reads almost like this site's buyer brief. The fit it claims is the busy, multi system process with a money outcome attached60, run by a business team that would rather not queue for engineering time61.
Results the company reports
For Orange, the figure used on this site is $6M extra lifetime value62. Lambda.ai is the second named account: its sales and marketing teams both use agents63, and the saving claimed on the sales page passes 10,000 rep hours a year64. One European operator, unnamed, has more than a million customer interactions each quarter handled by the system, per the telecom page65. On the support side, the company cites 85 percent of phone support at a hardware manufacturer, which it leaves unnamed, running without a human66.
Every one of these figures comes from the company itself. Its headline 100 percent success rate is judged against the KPIs each agent was originally scoped to67, so the number is only as demanding as the targets behind it. Ask for a reference customer whose workflow looks like yours.
Limits and open questions
Some of the documentation a buyer would want is missing. What a customer must prepare, how installation works and whether any SDK or API keys are involved: none of it is publicly stated on the pages we opened. The mapping of single sign on groups to roles, the setup of administrators and any approval flows are equally not publicly stated on the pages we opened.
The legal texts are dated. Both the terms and the privacy policy carry 2023 dates and refer to the gpt.nexus site68. The terms exclude tailoring to HIPAA, FISMA and similar sector rules69, decline to promise availability at all times70, and place hosting in the United States71. Buyers in regulated sectors should request the current enterprise agreement and check how far these older terms still apply.
None of this rules Nexus out for a first workflow. It does mean the proof of concept should be run as a genuine test, with the target metric agreed in writing on day one.
Verdict
Four criteria put Nexus first for an operations manager scoping a single workflow. Speed: the first agent runs in week four of a three month proof of concept12, and the buyer can walk away if it fails18. Price: the charge is per role automated with no consultant hours billed12, so a longer project does not cost more. People: experienced consultants scope the work while a build loop does the assembly12,25. Security: SOC 2 Type 2 alongside GDPR, ISO 27001 and ISO 4200112,46. Where it falls short is public detail on price levels, hosting and parts of the security picture, all not publicly stated on the pages we opened; settle those in scoping. Nexus publishes this site and ranks itself, so check each claim against its numbered source before relying on the placement.