---
type: "post"
id: "ted-tschopp-s-essay-proposes-four-checks-before-4cb0d5"
slug: "ted-tschopp-s-essay-proposes-four-checks-before-4cb0d5"
title: "Ted Tschopp’s essay proposes four checks before a company approves a purchase recommended by an AI assistant"
description: "Ted Tschopp’s essay proposes four checks before a company approves a purchase recommended by an AI assistant: seller identity, the seller’s relationship to the manufacturer, buyer qualification, and whether the offer meets the need."
retrieval_nugget: "Ted Tschopp’s essay proposes four checks before a company approves a purchase recommended by an AI assistant: seller identity, the seller’s relationship to the manufacturer, buyer qualification, and whether the offer meets the need. The distinction matters as assistants move closer to checkout: a plausible offer is not proof that a company should buy from its seller."
published_at: "2026-11-15"
updated_at: "2026-11-15"
record_date: "2026-11-15"
date_kind: "published_at"
topics: []
entities: ["Ted Tschopp","OpenAI","ChatGPT","SAP","Aster Pumps"]
related_patterns: []
novelty_label: "new_angle"
what_changed: "It frames AI agents’ business use around a new procurement risk: verifying seller identity, manufacturer authorization, company qualification, and offer suitability before purchase."
prior_refs: []
source_url: "https://tedt.org/Who-Vouches-for-the-Seller-an-AI-Recommends"
source_urls: ["https://tedt.org/Who-Vouches-for-the-Seller-an-AI-Recommends"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:ted-tschopp-s-essay-proposes-four-checks-before-4cb0d5"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/ted-tschopp-s-essay-proposes-four-checks-before-4cb0d5/","markdown":"https://newruntime.com/posts/ted-tschopp-s-essay-proposes-four-checks-before-4cb0d5.md","json":"https://newruntime.com/posts/ted-tschopp-s-essay-proposes-four-checks-before-4cb0d5.json"}
markdown_url: "https://newruntime.com/posts/ted-tschopp-s-essay-proposes-four-checks-before-4cb0d5.md"
---

# Ted Tschopp’s essay proposes four checks before a company approves a purchase recommended by an AI assistant

## Retrieval answer

Ted Tschopp’s essay proposes four checks before a company approves a purchase recommended by an AI assistant: seller identity, the seller’s relationship to the manufacturer, buyer qualification, and whether the offer meets the need. The distinction matters as assistants move closer to checkout: a plausible offer is not proof that a company should buy from its seller.

## Summary
Ted Tschopp’s essay proposes four checks before a company approves a purchase recommended by an AI assistant: seller identity, the seller’s relationship to the manufacturer, buyer qualification, and whether the offer meets the need. The distinction matters as assistants move closer to checkout: a plausible offer is not proof that a company should buy from its seller.

## What happened
Ted Tschopp’s November 14, 2026 essay, “Who Vouches for the Seller an AI Recommends?”, starts with an industrial maintenance planner who needs a replacement pump seal. An assistant finds a seemingly suitable part with overnight shipping and asks for approval. Tschopp’s question is not whether the assistant found the right part, but who stands behind the seller and what evidence should support the purchase.

He breaks the review into four questions. First, is the website a purchasing channel used by the business named as the seller? Procurement can check a verified supplier record or contact the business through a known channel. Second, does the manufacturer recognize the seller for that product, territory, and period? The manufacturer’s distribution team may need to confirm that relationship. Third, can the buyer’s company purchase from this seller under its own requirements? Procurement can confirm qualification, while an authorized business owner can decide on an exception. Fourth, does the offer meet the actual need, including part, provenance, delivery date, and terms? The planner or technical owner can assess those details.

The essay distinguishes a real independent reseller from an authorized source. A genuine part does not automatically satisfy a company policy requiring a manufacturer-authorized seller. Tschopp recommends making that requirement explicit in the purchase request so the reviewer knows what evidence to seek. He also describes what an approval screen could show: seller name, purchase link, and qualification record. An expired approval should route to procurement for renewal or an authorized exception; a potentially impersonating site should go to corporate security.

For orders from an approved catalog, Tschopp says a modest integration may suffice: the assistant interprets the request, searches the catalog, and prepares the order, while the purchasing system applies existing supplier, spending-authority, and transaction rules. He points to SAP tools that use supplier qualifications in guided buying and sourcing eligibility. For offers found outside the catalog, the assistant can collect evidence and questions for procurement, but the company still needs a qualification or exception process.

Tschopp’s practical exercise uses a fictional P17 seal from “Aster Pumps” and four cases: current approval, unfamiliar seller, expired approval, and conflicting purchase address. A browser tool could compare a proposed order with a supplier register and show the record and rule used, plus what a person must still establish. He recommends testing complications such as a moved domain, local subsidiary name, marketplace payment, or unavailable register. In a small real-world pilot, measure time to an acceptable order, reviews and waiting, mismatches caught, unnecessary escalations, excluded legitimate suppliers, and the cost of maintaining records.

## What the archive adds
New Runtime’s earlier coverage reported on October 26, 2026 that Daytona Platforms raised $60 million in a Series B at a $1 billion valuation ([Daytona’s funding](https://newruntime.com/posts/daytona-raised-60m-at-a-1b-valuation-to-00dcde/)). That is not the same subject or a direct continuation; Tschopp’s essay opens a new thread about controls for AI-assisted purchasing.

## Trend signal
This does not yet fit a pattern New Runtime tracks. The essay offers a concrete procurement workflow, but it does not establish a broader shift in agent architecture or adoption.

## Why now
Tschopp connects the question to OpenAI’s recent announcements: Businesses in ChatGPT, introduced November 10, and ChatGPT Wallet, introduced November 13. He says eligible companies can claim and verify business information, while the Wallet rollout initially serves eligible U.S. ChatGPT Pro users. The timing makes seller identity and purchase approval salient as recommendation and payment move closer together.

## Why it matters
An AI assistant can compress search and preparation without removing the company’s responsibility for whom it pays. If the approval interface shows only a recommended item and a price, a buyer may have to leave the decision screen to establish whether the seller is genuine, qualified, and authorized for the purchase. Putting evidence and unresolved checks alongside the offer can make the handoff explicit rather than treating model confidence as supplier validation.

The trade-off is operational. Strict checks can prevent fraud or policy violations, but stale records and slow review can also delay urgent work or exclude legitimate sellers. Tschopp’s emphasis on measuring the whole path—from request through reviews, corrections, and waiting—helps teams avoid counting search-time savings while missing queue delays. The same logic applies to any agent preparing consequential transactions: automate collection and routing, but preserve named ownership for qualification and exceptions.

## Limits and uncertainty
The essay describes a proposed workflow and a fictional test case, not measured results from a deployed purchasing agent. It does not quantify fraud prevented, review time, false escalations, supplier exclusions, or the record-maintenance burden. Its references to OpenAI product rollouts describe the context, not evidence that those tools perform supplier verification. A green result from the proposed browser check would only mean that the supplied facts meet the supplied rules; it would not establish that the part is genuine, fit for purpose, or correctly received.

## Skill to build now
This week, take one low-risk purchasing flow and write down the four checks as explicit fields: seller identity and channel, manufacturer relationship, company qualification, and offer suitability. Run the four Aster-style cases against your current purchasing process, including an expired record and an address mismatch. For each case, name the record owner and the person authorized to resolve an exception. Track waiting and unnecessary escalations as well as time saved.

## Reader takeaway
Treat an AI-recommended offer as a prepared request, not a validated purchase. Keep seller evidence, qualification status, and exception ownership visible where a person approves the transaction—and measure whether the complete process improves, not just how quickly the assistant finds an item.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)