Companies continue to approve large AI budgets while few projects reach reliable daily operation. Insight Partners has formed a practice that places engineers inside portfolio companies to close that gap. The move treats the shortage of staff who can integrate models into existing systems as the binding constraint on progress.
The prior pattern
Most enterprise AI work has stayed in pilot form. Teams train models, run proofs of concept on limited data, and produce accuracy metrics that look promising in isolation. When the same models encounter live data pipelines, legacy access controls, and shifting business rules, performance drops and maintenance costs rise. Few organizations have kept the original pilot teams intact long enough to solve those integration problems at scale.
Insight Partners observed this pattern across its holdings and decided the missing piece was not better model research but engineers who stay on site until the system runs without constant hand-holding. Operating Partner Pablo Dominguez described the new practice on Bloomberg Tech as an attempt to build repeatable internal capability rather than deliver one-off projects.
How the practice is structured
The firm is assembling teams of forward-deployed engineers who pair model work with the concrete requirements of production environments. That includes building or repairing data pipelines, adding monitoring that surfaces drift or cost spikes, and connecting outputs to the workflows already used by sales, operations, or finance teams. Dominguez noted that the firm sees the translation step—turning a working notebook into a service that meets uptime, audit, and latency targets—as the skill most in short supply.
No customer names or contract values were disclosed during the segment. The emphasis instead fell on repeatability: once a portfolio company completes one successful deployment, the same engineers and processes can be applied to the next use case without starting from scratch each time. The practice therefore functions as both a delivery mechanism and a training program for the companies involved.
Limits of the current evidence
The Bloomberg segment contained no data on how many companies have already used the new practice or what fraction of their AI projects have moved from pilot to production. It also did not address whether the embedded engineers report to Insight, to the portfolio company, or to a joint steering group. Those governance details will determine whether the approach reduces coordination overhead or simply adds another layer of stakeholders.
Why it matters
Enterprises that approved AI spending in prior budget cycles now face pressure to show operating results before the next planning round. When models remain in staging, finance teams have an easy argument for cutting future allocations. Forward-deployed engineers who remain accountable until systems stabilize offer one concrete way to produce measurable output rather than another round of slide decks.
For engineering organizations already inside those companies, the arrival of external specialists with direct authority over integration decisions changes day-to-day work. The specialists bring experience from other deployments, which can shorten trial-and-error cycles. At the same time, they can override local priorities if their mandate is defined too broadly. The outcome will depend on whether scope, decision rights, and exit criteria are written down at the start of each engagement.
The larger signal is that the bottleneck after model training has shifted from raw technical performance to organizational execution. Insight’s response is to supply the missing execution capacity directly rather than wait for portfolio companies to hire it on their own. That choice reveals where the firm believes the next two years of value will be created or lost.
---
Sources:
{
"sources": [
{
"publisher": "Bloomberg Technology",
"title": "AI Moves From Experiment to the Enterprise",
"url": "https://www.bloomberg.com/news/videos/2026-10-05/ai-moves-from-experiment-to-the-enterprise-video",
"published_at": "2026-10-05T17:37:36.000Z",
"summary": "Companies are spending billions on AI, but actually getting it into production remains a major challenge. Insight Partners sees forward-deployed engineers as part of the solution: embedding directly with customers to turn AI into working products and workflows. Insight Partners Operating Partner Pablo Dominguez joins to discuss why the firm is launching a dedicated practice to help its portfolio companies build that capability. He joins Ed Ludlow on \"Bloomberg Tech.\""
}
]
}
No comments yet