Google I/O 2023: artificial intelligence moves from experiments into work tools

On 10 May 2023, Google used its I/O conference to present a strategy in which generative AI was no longer isolated in a laboratory. Search, Workspace, Cloud and developer tools were beginning to incorporate generation and assistance.

The announcement clarified the next stage of digital transformation. AI would arrive not only as a standalone application, but inside the tools where people already search, write, analyse and collaborate.

Search begins to change

Google introduced the Search Generative Experience, or SGE, as a Search Labs experiment. It could assemble a generative response, present perspectives and allow follow-up questions.

For users, this meant moving from a short query towards conversational exploration. For companies and publishers, it raised questions about visibility, sources, clicks and advertising. Search would still need accessible web pages, but it could reorganise their information before a visit.

AI inside everyday work

Google also demonstrated generative features for drafting, summarising, organising and creating. The promise was to reduce time spent on a first version and allow people to focus on judgement and decisions.

Yet faster drafting does not automatically improve an outcome. A fluent document may contain incorrect facts, a summary may omit a condition and a prompt may expose information that should remain private. Productivity appears when the organisation defines which tasks may be assisted, which sources are valid and who approves the result.

From documents to governed data

Generative AI exposes a familiar problem: business information is scattered across email, spreadsheets, documents and applications. If each source contains a different version, an assistant cannot know which one should prevail.

Companies therefore need an authoritative source, access rules, document lifecycle, change records and a policy for what may be used as model context. This work improves ecommerce, analytics and customer service as well as AI.

Starting with discipline

A first implementation should select a repetitive, low-risk task and measure time, correction rate, common errors, user satisfaction, tool cost and impact on the complete process. If there is no demonstrated improvement, adding more AI only adds complexity.

Sitelicon connects strategy, technology, data and operations to turn AI capabilities into useful services. The lesson from May 2023 remains relevant: access to a tool can be immediate, but integrating it well requires priorities, architecture and learning.

Editorial note: originally published in May 2023 and reviewed on 11 September 2026 to preserve the continuity of our editorial archive.