AI is finally giving us the intranet we were promised
In November 1998, two equity analysts at Merrill Lynch described the future of work in a single sentence. Christopher Shilakes and Julie Tylman were not futurists or vendors with something to sell.
They were a research desk putting a number on a market that did not yet have a name — and they coined the name in the same report: the Enterprise Information Portal. An EIP, they wrote, would give a company’s people “a single gateway to personalized information needed to make informed business decisions.”
That phrase outlined much of the promise of corporate AI systems. The full note describes the thing every organization still wishes it had: one place, personal to the user, where the company’s information comes together. They believed the market for it would rival or exceed all of ERP; bigger than SAP, bigger than PeopleSoft. If you look at our AI incumbents they were right, it just took longer than they thought.
The goal has always been clear; the technology was simply harder than people thought. The idea sounds simple: get all the company’s data in one place and let non-technical people process the data to deliver useful insights. If you have seen an ERP project play out, you know it was never going to be easy.
How it started
For twenty-seven years the industry threw a new solution architecture at the problem roughly every five years. Each evolution solved a real problem but none reached the goal. Each time they were the ceiling of what was buildable.
ERP attempted to answer fragmentation with consolidation. One system, one schema, nothing left to integrate. It simply couldn’t keep up. It created monolithic systems that were very technically intensive to set up and maintain. In the end the costs were higher than most organizations could afford and best-of-breed SaaS multiplied faster than any single system could evolve.
Integration middleware accepted the fast-moving SaaS landscape and took the opposite and equally reasonable path. Leave the systems separate and wire them together. While theoretically simpler, the solution was very complex and brittle. Every data seam had to be hand-built and maintained as disparate systems rapidly evolved.
The enterprise portal built in SharePoint and its kind delivered some of the benefits but ended up being implemented more like a directory. In attempting to bring one solution to every business, the learning curve was daunting and few ever really took off with users, in most cases devolving into shared file systems.
The data warehouse created unified data for analysis by enabling a read-only, after-the-fact view of the business. These could often deliver great insights but the technical complexity required real expertise to take full advantage of their potential, limiting use to a select few in an organization.
What changed
AI happened. Along with MCP (Model Context Protocol) and the embrace of data APIs, it changed the equation completely. Current AI systems are inherently data engineers, given context and access they can pull from HubSpot, a company directory, a data marketplace, and proprietary data in a single conversation. This bypasses many of the normalization challenges that encumbered every previous approach. The seams that required integration specialists to hand-build and maintain simply matter less now. Non-technical users can access and use data across systems without a data team standing between them and the answer.
The chat interface solves the second big issue by making the complexity manageable for non-technical staff. Sales leaders can chat and structure reports. Finance leaders can question directories and interrogate contract terms. This eases the burden for much of this work, though it is critical to remember that all AI output requires human review and cannot be blindly accepted. The data gap is closing, and the promise of the intranet expressed by Shilakes and Tylman in 1998 looks like it might finally become a daily reality.
Where we are now
So how does this actually happen and what does it look like? It starts with a portal, a governed safe space where data moves freely, but only on the organization’s terms. Two pillars hold the boundary; two layers do the work. The most advanced organizations add a third layer that lets individuals and teams build and host their own tools inside that safe space, never outside it.
What this provides is flexibility and speed. Any person in the organization, via a simple chat interface, can access the data their role allows and get the answers they need when they need them. That increases the velocity of the entire organization.
This stack would fulfill the promise of that 1998 paper, “a single gateway to personalized information needed to make informed business decisions.” Just with tools Shilakes and Tylman could not have imagined.
On the 1998 Paper
The report mentioned is Enterprise Information Portals, by Christopher C. Shilakes and Julie Tylman, Merrill Lynch & Co. Securities Research, dated November 16, 1998.