INNOVAREModule 10 · Managing & Exploring AI Capabilities and Adoption

Case 3: The Absorption Problem

Cohen & Levinthal's absorptive capacity — the module's second theory lens — taught through the half-billion-dollar algorithm that couldn't be questioned.

August 2026 · Case 3 of 6
As you read — hold this question

A firm can buy the best AI in its market. Why is that not the same as being able to use it?

$500M+
Zillow's home-buying algorithm kept paying yesterday's prices as the market turned. Zillow shut the unit down in Nov 2021, wrote off more than half a billion dollars and cut ~25% of staff.

The module's second lens is absorptive capacity: recognise, assimilate and apply new knowledge. Zillow had the data (acquisition) but couldn't assimilate what its black-box model was doing or transform its process when the world moved — the exact failure the theory predicts.

The second theory lens
The textbook versionQuiz: absorptive capacity

Absorptive capacity (Cohen & Levinthal; Zahra & George)

Absorptive capacity (Cohen & Levinthal, 1990): a firm's ability to recognise, assimilate and apply new external knowledge — and it depends on prior related knowledge. Zahra & George (2002) split it into Potential (Acquisition, Assimilation) and Realized (Transformation, Exploitation), leading to competitive advantage (flexibility, innovation, performance).

Two corrections the deck needs
The slide dates Cohen & Levinthal as β€œ2024” — it is 1990. And the Potential/Realized box diagram it shows is actually Zahra & George (2002), not Cohen & Levinthal.
Carry this

Acquisition & Assimilation = Potential; Transformation & Exploitation = Realized. Buying knowledge is Potential; using it is Realized.

The half-billion-dollar failure
The rest of the story

Zillow, read as an absorptive-capacity story

DimensionZillow Offers
Acquisition✓ Best housing dataset in the US.
Assimilation✗ Executives couldn't understand the black-box model's failure modes.
Transformation✗ Automated buying kept buying while human selling slowed — process couldn't adapt.
Exploitation✗ Exploited outputs into negative value — overpaying for homes.

Barton's own words on the shutdown: β€œthe unpredictability in forecasting home prices far exceeds what we anticipated.”

Why the theory needs an update
The other side

The GenAI inversion — and the black box

Classic absorptive capacity assumed absorbing technology required deep technical expertise. Natural-language GenAI inverts this: any employee can exploit the output instantly — but the firm may never assimilate the underlying knowledge, building a hollow, vendor-dependent capability. And the standing critique (Lane et al., 2006) is that the construct became a black box — which Zillow made literal.

The other side

The 2026 bottleneck is no longer individual technical literacy — it is structural: data permissions, process redesign, and whether anyone can actually question the model.

Quick recall — without looking back

Test yourself on this case

Question 1 of 3

Define absorptive capacity and its two-stage (Zahra & George) structure.

A firm's ability to recognise, assimilate and apply new knowledge (Cohen & Levinthal, 1990). Zahra & George (2002): Potential (Acquisition, Assimilation) → Realized (Transformation, Exploitation).
Question 2 of 3

Which absorptive-capacity dimensions did Zillow fail, and how?

It had Acquisition (data) but failed Assimilation (couldn't understand the black box), Transformation (process couldn't adapt) and Exploitation (created negative value) — >$500M write-off.
Question 3 of 3

Name the two citation errors in the deck's absorptive-capacity slide.

Cohen & Levinthal is dated 1990, not 2024; and the Potential/Realized diagram is Zahra & George (2002), not Cohen & Levinthal.

Module 10 Videos

Module 10 · Long Form

Sources

Module content
BUSN9049 Module 10 — absorptive capacity (deck + videos). Flinders University, 2026.
Origin & critique
Cohen & Levinthal, ASQ (1990); Zahra & George, AMR (2002); Lane, Koka & Pathak, AMR (2006).
Case
Zillow Offers shutdown, Nov 2021 (Q3 2021 earnings; ~2,000 layoffs / 25%; >$500M write-down).